{"companyTimeline":[{"id":"timeline_01","date":"2026","title":"Multiple Financing Rounds Completed","summary":"Pre-A++ round of financing has been successfully concluded, drawing joint investments from funds affiliated with Legend Holdings, Fullhan Microelectronics, Zhejiang Venture Capital, JinkoSolar, among other institutions, in addition to sustained follow-on investments from existing shareholders.","summaryHtml":"","sortOrder":6,"isActive":true,"textStyles":{}},{"id":"timeline_02","date":"2025","title":"100+ Million in Pre-A+ Financing","summary":"Secured 100+ million in Pre-A+ financing from investors including Hubble Investment.","summaryHtml":"","sortOrder":5,"isActive":true,"textStyles":{}},{"id":"timeline_03","date":"2024","title":"Global Expansion","summary":"KOKONI products have reached more than 20 countries and regions.","summaryHtml":"","sortOrder":4,"isActive":true,"textStyles":{}},{"id":"timeline_04","date":"2023","title":"Over 100,000 Users","summary":"The product user base surpassed 100,000.","summaryHtml":"","sortOrder":3,"isActive":true,"textStyles":{}},{"id":"timeline_ff45128b-1de9-4e77-a889-ebf2933d1e44","date":"2022","title":"Pre-A Financing","summary":"Revenue exceeds 10 million,wth Pre-A round financing amounting to tens of millions of yuan led by Deshi Investment.","summaryHtml":"","sortOrder":2,"isActive":true,"textStyles":{}},{"id":"timeline_1782699546807_ffyf43r7","date":"2021","title":"Company Founded","summary":"KOKONI was founded and completed its angel financing round.","summaryHtml":"","sortOrder":1,"isActive":true,"textStyles":{}}],"companyNews":[{"id":"news_e8e971d6-b565-4400-8511-39afa66ac63b","date":"2026-11-04","time":"2026.11.04","title":"MoWorld 2.0 4D World Model Debuts at the Wuzhen Summit","summary":"Helping AI truly understand the physical world","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/companyNews/1794203971068-0fe3f92d-6b41-4c08-9860-a5bf8b3f627b.png","mediaType":"image","sortOrder":1,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_651a4b2b-3e51-409b-bae5-049daecd9e5b","type":"paragraph","subheading":"","html":"<p>On November 4, at the World Internet Conference forum on AI as a driver of new productive forces and growth, KOKONI founder and CEO Tianrun Chen joined representatives from three national AI application pilot bases—in embodied intelligence, culture and tourism, and spatiotemporal geographic information—to officially launch MoWorld 2.0, a 4D world model built natively on Huawei Ascend.</p><p>Following the release of MoWorld 1.0 in July, MoWorld 2.0 delivers another leap in model capabilities. It incorporates gravity, elasticity, buoyancy, lighting and heat, as well as the characteristics of rigid bodies, soft bodies, fluids and deformable bodies, into spatial modeling and interaction. Its output consists of reusable 4D Gaussian assets that support simulation and real-time rendering.</p>","align":"left","textStyles":{}},{"id":"block_6cc0f6a1-b466-4991-b4c3-102a469c251d","type":"paragraph","subheading":"","html":"<p>From dynamic scene understanding and physical interaction to real-time inference and spatial asset generation, MoWorld 2.0 is expanding what world models can do: it builds dynamic worlds that people can enter, interact with and change, going beyond image generation.</p><p>As cooperation with the three national AI application pilot bases deepens, we will accelerate real-world deployment and wider adoption in interactive simulation and embodied training, cultural digital assets and immersive experiences, real-scene 3D data production, and rapid reconstruction of large environments.</p><p>“We want AI to establish accurate, comprehensive and deep connections between space, time and physical laws, beyond depicting how the world looks,” Chen said at the forum. “We are confident that combining domestic computing power with world models will make AI a new productive force that truly understands the physical world and creates more possibilities for the future.”</p>","align":"left","textStyles":{}},{"id":"block_067f173e-55a0-4899-a36b-d9f7e9553875","type":"image","src":"/uploads/companyNews/1794205740440-e4af08b5-c627-4171-a317-db9c49f59ec5.jpg","alt":"KOKONI founder and CEO Tianrun Chen launches MoWorld 2.0 at the forum","caption":"KOKONI founder and CEO Tianrun Chen launches MoWorld 2.0 at the forum","layout":"normal"},{"id":"block_1ee188c2-3090-4046-b78b-e2fb7c34a4b2","type":"paragraph","subheading":"From Spatial Reconstruction to Physical Modeling","html":"<p>While video generation primarily addresses what a scene looks like, MoWorld 2.0 reasons about a further question: how will the world change when an action occurs?</p><p>By modeling multiple physical factors, MoWorld 2.0 generates spatial and object appearances while also reasoning in real time about interactions, external forces and environmental changes.</p><p>Simulating a breeze turning a page and a bookmark falling to the floor, for example, requires considering wind strength and direction, the flexibility of paper, the bookmark&#x27;s weight, and the rigidity and contact relationships of the floor. A seemingly simple change involves several physical factors working together.</p>","align":"left","textStyles":{}},{"id":"block_3b8bd015-79b2-4c75-abe9-c985053a71b9","type":"image","src":"/uploads/companyNews/1794204715748-cddd6110-9d4e-4893-8496-bd1099307758.gif","alt":"Simulating a breeze turning book pages","caption":"Simulating a breeze turning book pages","layout":"normal"},{"id":"block_7267965a-93d2-4dc3-b1a5-375e076e4853","type":"image","src":"/uploads/companyNews/1794204774768-21e1625e-355b-4049-b5b4-df9fd9bc2159.gif","alt":"Simulating a bookmark falling to the floor","caption":"Simulating a bookmark falling to the floor","layout":"normal"},{"id":"block_faea735d-a858-475c-8b3b-719dd811c248","type":"image","src":"/uploads/companyNews/1794204806646-e5ebca39-e131-40b5-a1d4-1d89b39bbca1.gif","alt":"Editable, interactive spaces","caption":"Editable, interactive spaces","layout":"normal"},{"id":"block_8e9af2b0-2726-474e-a6b6-897ab5c24e0d","type":"image","src":"/uploads/companyNews/1794204844473-9414eb7c-ba91-4205-9adb-534c50cee855.gif","alt":"Edit space and adjust the pace of time","caption":"Edit space and adjust the pace of time","layout":"normal"},{"id":"block_ec4eb26f-03d7-4d3f-b7eb-7fffec78c348","type":"paragraph","subheading":"","html":"<p>MoWorld 2.0 brings these factors into a unified spatial interaction process, allowing actions to produce corresponding effects and turning scenes into more than static backdrops.</p>","align":"left","textStyles":{}},{"id":"block_da91f075-fba3-4a16-9412-5747e0a15f58","type":"video","src":"/uploads/companyNews/1794204960470-da6bf51f-0131-4b37-87c6-2e22b22c9e63.mp4","poster":"","caption":"","layout":"normal"},{"id":"block_5714ce99-9ac5-499e-936d-1b28d8d89c51","type":"paragraph","subheading":"","html":"<p>Our goal is to move world models from static spatial reconstruction to modeling dynamic physical worlds, and from producing continuous images to generating reusable 4D assets.</p><p>Generated spaces can be retained, objects edited and scenes called upon repeatedly. Users can enter a space to observe it, change its contents and state, and reuse it for later creation, training and validation.</p><p>Globally, understanding and integrating these physical factors as a whole represents a leading stage in the transition from static spatial reconstruction to dynamic physical world modeling, and from spatial imagery to 4D asset output.</p><p>“This will make generated spaces feel more interactive and realistic,” Chen said. “The model will build a digital world that people can enter, interact with and change.”</p>","align":"left","textStyles":{}},{"id":"block_d0b6b48e-a0b6-4163-a407-0f85cbda6e32","type":"paragraph","subheading":"Faster Real-time Inference and More Applications","html":"<p>MoWorld 2.0 also improves real-time performance:</p><p>■ Video inference: increased from a peak of about 50 frames per second in version 1.0 to about 70 frames per second.</p><p>■ Spatial asset rendering: 3D/4D Gaussian assets can be rendered at more than 500 frames per second.</p><p>■ Mobile rendering: related tests have reached about 100 frames per second.</p><p>Higher real-time performance provides faster feedback to user actions and a stronger engineering foundation for deployment on phones, robots, vehicles and industrial equipment.</p><p>In October, KOKONI&#x27;s “Mo Xiao Chong” pet app appeared at a Huawei smartphone launch, offering another example of MoWorld reaching consumer devices. Users upload a few pet photos to create a 4D digital pet with spatial structure that can be viewed and interacted with from different angles.</p><p>An end-to-end device-and-cloud pipeline makes this possible. MoWorld 2.0 is trained on Huawei Ascend and compressed into an on-device version with around 600 million parameters for corresponding inference on the Kirin platform. The cloud generates high-quality spatial assets, while the phone handles rendering, interaction and presentation.</p><p>MoWorld is also being deployed in embodied intelligence training, autonomous driving, industrial digital twins, film and entertainment production, and culture and tourism. KOKONI is among the few companies globally to engineer world model technology into deployable systems and deliver it at scale.</p><p>This is only the beginning.</p><p>KOKONI and its partners are progressing through application validation, implementation and product delivery in embodied intelligence training, autonomous driving, industrial digital twins, film and entertainment production, and culture and tourism.</p><p>For embodied intelligence training, MoWorld 2.0 builds virtual environments with spatial structure and basic physical feedback. Robots can train, experiment and review their actions repeatedly, allowing some exploration to take place in digital space first.</p>","align":"left","textStyles":{}},{"id":"block_6aed11ae-cc6c-4df9-b8e7-98e59600c881","type":"image","src":"/uploads/companyNews/1794205218224-465d9604-f59c-4942-a742-001c2ea12bdd.gif","alt":"","caption":"","layout":"normal"},{"id":"block_9ccb60b1-683b-4b38-9a33-d95ae5c6c9e1","type":"paragraph","subheading":"","html":"<p>For autonomous driving training, the model generates long-tail scenarios—rain and snow, potholes, unexpected obstacles and complex traffic—in a unified 3D/4D space. This expands the environments available for repeatable testing and improves training-data production efficiency.</p>","align":"left","textStyles":{}},{"id":"block_10ce598f-fb8b-4bd9-b48a-a15d71491b92","type":"image","src":"/uploads/companyNews/1794205276789-d9d02b82-90af-4b37-a444-60ea97760f23.gif","alt":"","caption":"","layout":"normal"},{"id":"block_56b7e409-2f81-4e8c-90e5-4b5a43aaf89a","type":"paragraph","subheading":"","html":"<p>For industrial digital twins, real environments such as factory production lines and mines become editable spatial assets for equipment layout changes, production-line modification rehearsals, remote collaboration, abnormal-scenario simulation and digital training.</p><p>In urban and rural planning, culture and tourism, these capabilities support spatial proposals, virtual tours, digital attractions and immersive content, making spaces easier to communicate and experiences easier to access.</p><p>In film and games, reusable scene assets support free camera movement, object editing and weather changes. They reduce repeated set construction and shorten storyboard previsualization and proof-of-concept cycles, giving creators more time for their ideas.</p><p>From consumer experiences to professional production, MoWorld is enabling repeated use of the same space and bringing model capabilities into a critical stage of industry adoption.</p>","align":"left","textStyles":{}},{"id":"block_285cebcf-6022-4d5c-9911-f8d7e6c70156","type":"paragraph","subheading":"Continued Innovation Supported by Domestic Computing and Industrial Capital","html":"<p>MoWorld 2.0 continues to be trained and deployed entirely on domestic Ascend infrastructure. Joint innovation with Huawei teams focuses on long-sequence training, coordinated scheduling across multiple NPUs and inference-system optimization to improve training efficiency, inference speed and deployment. According to company tests and estimates, inference costs can be as low as 30% of a GPU solution of comparable scale.</p><p>Lower costs affect whether world models can run continuously across more scenarios and reach more people and businesses, as well as improving engineering metrics. Close collaboration between domestic computing platforms and model development is opening new opportunities for deployment at scale.</p><p>Industrial capital continues to support this direction. This year, KOKONI has completed five financing rounds totaling RMB 2 billion. Investors include state-backed institutions, industrial investors and technology investment firms from China and abroad, covering chips, computing, smart devices, content platforms and advanced manufacturing.</p><p>This support will continue to fund model R&amp;D, domestic compute collaboration, data systems and industry applications, turning research into products that can be replicated and delivered sustainably.</p><p>The release of MoWorld 2.0 is both a model upgrade and an important step toward KOKONI becoming infrastructure for the spatial intelligence industry.</p><p>Model capabilities, domestic computing and industry applications are progressing together. We look forward to working with more partners to move world models from generation toward interaction, deployment and practical use.</p><p>MoWorld 2.0: reach what you can imagine, and what lies beyond your sight.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_394d4977-38f2-4adc-aec4-e0f4a4e6343d","date":"2026-09-04","time":"2026.09.04","title":"Governor Liu Jie Visits KOKONI, Calling for Stronger Support and Closer Integration of Innovation and Industry","summary":"Zhejiang Governor and Deputy Party Secretary Liu Jie recently visited KOKONI for a focused research visit.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/companyNews/1794203308882-d6ddbec3-bb11-4c8b-842f-0943475267fb.png","mediaType":"image","sortOrder":2,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_31fcf644-d91a-498d-931e-7c68c2c3d330","type":"paragraph","subheading":"","html":"<p>On September 3, Zhejiang Governor and Deputy Party Secretary Liu Jie visited KOKONI for a focused research visit. After hearing about progress in core technology R&amp;D and application development and watching a MoWorld 4D demonstration, he encouraged the company to keep upgrading its models, actively explore new business models and continuously improve the quality and efficiency of its development.</p><p>During the visit, Liu watched a demonstration of digital spaces generated by MoWorld 4D and practical exercises by robots trained in model-generated environments. He stressed the importance of opening more high-value scenarios to support scaled, commercial AI applications, strengthening resource support, attracting more long-term and patient capital, coordinating the entire industry chain—from chips and large models to agents and application services—and accelerating the deployment of innovation so AI can serve more industries.</p>","align":"left","textStyles":{}},{"id":"block_c8b28c6b-0675-4d0b-8b16-7375c4c1a6f9","type":"image","src":"/uploads/companyNews/1794203345578-f64cd4c8-3fbb-4055-bbd9-8ae5425c90bb.png","alt":"Governor Liu Jie hears KOKONI founder Tianrun Chen describe MoWorld 4D applications in robot training","caption":"Governor Liu Jie hears KOKONI founder Tianrun Chen describe MoWorld 4D applications in robot training","layout":"normal"},{"id":"block_6da397f1-780e-4d31-a7b7-0f92f76ee4b9","type":"paragraph","subheading":"Full Report","html":"","align":"left","textStyles":{}},{"id":"block_2fd348f4-5319-479c-98fe-4a7883dd0ecc","type":"image","src":"/uploads/companyNews/1794203403001-9843b85e-790e-4a19-a218-96e03609caa6.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_d5c9732b-29cd-46f4-9860-242126f9aca8","type":"video","src":"/uploads/companyNews/1794203483319-c2ffbc52-2283-4813-bc1a-9725e0dbe99e.mp4","poster":"","caption":"","layout":"normal"},{"id":"block_3d26421e-564b-41f3-800f-d4dc696e0f2c","type":"paragraph","subheading":"","html":"<p>On September 3, Zhejiang Governor and Deputy Party Secretary Liu Jie conducted a focused research visit on AI industry development in Hangzhou. He called for studying and implementing Xi Jinping&#x27;s statements on AI, responding to technological and industrial change, advancing the “AI Plus” initiative, integrating innovation and industry more deeply, and promoting high-quality growth in AI-related industries to strengthen a leading center for AI innovation and development.</p><p>KOKONI is a high-tech company focused on independently developed general-purpose world models. Liu watched a model demonstration, heard updates on core technology R&amp;D and application development, and encouraged continued model upgrades, exploration of new business models and improvements in development quality and efficiency. At Jing&#x27;an Technology, which develops core technologies and AI products, he reviewed AI model development and urged closer integration with real scenarios and smarter products and solutions. At Youlu Robotics, which develops large models for embodied intelligence, he visited the laboratory to learn about research and deployment, encouraging stronger core technologies, more generalizable AI brains, better hardware-software compatibility, and greater market recognition.</p><p>Liu emphasized that the AI industry is evolving rapidly. He called for further development of general-purpose and domain-specific large models, faster AI agent adoption and high-quality datasets to support more industries. More high-value scenarios should be opened to enable commercial applications at scale, with coordinated development of chips, models, agents and application services and faster conversion of innovations into practical results. He also urged deeper integration of AI and manufacturing, support for technological upgrading and the digital transformation of physical industries, and progress toward green, intelligent and integrated development. Resource support, long-term and patient capital, coordinated compute services and computing-network-data-energy systems, intellectual property protection, and interdisciplinary talent with both industry and AI expertise would help build a first-class AI innovation ecosystem.</p><p>Du Xuliang joined the visit.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_152fb2bf-a066-46f9-8c31-6e1e419a55af","date":"2026-07-18","time":"2026.07.18","title":"Building Worlds, Shaping the Future: KOKONI Hosts the MoWorld 3D World Model Launch","summary":"MoWorld 3D launches on Huawei Cloud, alongside a joint initiative to build future digital city infrastructure.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/companyNews/1786931245806-d8cb7d5c-a05a-4f52-8ff7-3f1006bff3d6.png","mediaType":"image","sortOrder":3,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_520bb6bc-a1d5-4d79-8d9f-ea0278d64606","type":"paragraph","subheading":"","html":"<p>On July 16, KOKONI, a national AI application pilot base and Huawei Cloud jointly hosted “Building Worlds, Shaping the Future” in Hangzhou. The event marked the launch of MoWorld 3D on Huawei Cloud and the start of a joint initiative to build future digital city infrastructure, bringing world models closer to 3D spatial construction and industry applications. As the first world model launch since the national pilot platform was inaugurated, it showcased KOKONI&#x27;s latest progress in domestic computing, 3D world models and spatial intelligence applications.</p><p>Nearly 150 guests from industry, academia and investment attended. Participants included Gao Chong, a member of the Party Working Committee, deputy director of the administrative committee and executive deputy district head of Hangzhou High-tech Zone (Binjiang); representatives of Hangzhou&#x27;s Development and Reform Commission and Binjiang&#x27;s Development and Reform Bureau; and companies including Huawei Cloud, Transfar Group, Legend Holdings, Hyundai Motor, Geely, Hangkai Holdings and Huace Film &amp; TV.</p>","align":"left","textStyles":{}},{"id":"block_d1909c0d-0fd6-4ff2-badf-b7fe78f6fdfd","type":"video","src":"/uploads/companyNews/1786931283728-9ef07f40-dec5-40cc-ad6b-0cfd1c2ab617.mp4","poster":"","caption":"","layout":"normal"},{"id":"block_16e0d1b7-1f21-4720-b22f-5356b41a3c95","type":"paragraph","subheading":"Government and Industry Build a World Model Ecosystem","html":"<p>In his address, Gao Chong said Binjiang is working to become a leading district for AI industry innovation and development, with rapid, high-quality growth. In 2025, the district&#x27;s AI industry reached RMB 47.83 billion, accounting for 47.3% of Hangzhou&#x27;s total. Its national embodied intelligence application pilot base has also created an early advantage in embodied intelligence. He described MoWorld 3D as an original core technology developed through the base&#x27;s public innovation platform and Huawei Cloud computing resources, further strengthening the spatial intelligence technology system. Open source is a starting point and practical applications are essential, he said, calling on industry, academia, research institutions and users to accelerate exploration and bring solutions to more industries and households.</p>","align":"left","textStyles":{}},{"id":"block_2f5ea8eb-3e85-4f0e-b95b-b9a740cdb176","type":"image","src":"/uploads/companyNews/1786931341159-c8fa0f16-d0c3-45fe-a214-45f862480279.jpg","alt":"Gao Chong, executive deputy district head of Hangzhou High-tech Zone (Binjiang)","caption":"","layout":"normal"},{"id":"block_d188e482-d4d7-4366-a33a-93cf65cb332d","type":"paragraph","subheading":"","html":"<p>Yao Pengzhi, general manager of Huawei Cloud Zhejiang, said innovation relies on a thriving industrial ecosystem rather than isolated breakthroughs. MoWorld 3D demonstrates Huawei Ascend Cloud&#x27;s integrated, full-stack capabilities from training to inference. Ascend Cloud supports the model&#x27;s computing demands and offers a new route for AI to understand the complex physical world.</p>","align":"left","textStyles":{}},{"id":"block_0c6b6310-827e-4b64-b27e-8288fe50cc95","type":"image","src":"/uploads/companyNews/1786931377619-f09f5e0c-d82f-45b3-8ac7-dd617b6d1d4a.jpg","alt":"Yao Pengzhi, general manager of Huawei Cloud Zhejiang","caption":"","layout":"normal"},{"id":"block_62453be6-a73b-4132-bd41-609d2677ac82","type":"paragraph","subheading":"MoWorld 3D Launches: World Models Advance into 3D Spatial Construction","html":"<p>KOKONI founder and CEO Dr. Tianrun Chen officially introduced MoWorld 3D and explained its capabilities. He described the move from 2D visual perception to 3D spatial understanding as a crucial step toward AI engaging with the physical world, with MoWorld 3D offering a clear solution built on domestic technology.</p>","align":"left","textStyles":{}},{"id":"block_aaee9513-9f3b-4584-b5ca-b4a08074e1fd","type":"image","src":"/uploads/companyNews/1786931429129-76e9ebfd-e135-4d83-8906-72224b464738.jpg","alt":"KOKONI founder and CEO Dr. Tianrun Chen","caption":"","layout":"normal"},{"id":"block_fb8a4ceb-fe57-482a-913f-f8490765951f","type":"paragraph","subheading":"","html":"<p>Chen is a doctoral researcher at Zhejiang University&#x27;s College of Computer Science, supervised by Pan Yunhe, an academician of the Chinese Academy of Engineering. His research focuses on 4D modeling, large multimodal models and world models. Drawing on Zhejiang University&#x27;s CAD&amp;CG laboratory, KOKONI continues to translate world model research into industrial applications.</p><p>KOKONI previously released MoWorld, its first interactive world model with an ultra-high frame rate, and introduced the Flash World Model concept.</p><p>MoWorld generates worlds that respond in real time at high frame rates to images, text prompts and user actions. As users move forward, step back or change their view, it generates the next view based on camera actions. MoWorld 3D further changes how world models represent the physical world.</p>","align":"left","textStyles":{}},{"id":"block_87a4d79d-85c8-4b4b-b9e4-fa606a3cda04","type":"image","src":"/uploads/companyNews/1786931518159-bde4f207-4b2b-4659-acc5-bce3e4d1aa79.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_18bdcb29-08e4-4b0f-8f54-b286cb5ab657","type":"paragraph","subheading":"","html":"<p>MoWorld 3D goes beyond generating what is visible from a single viewpoint. It unifies geometry, semantics and dynamics in 3D/4D space through camera-pose estimation, depth and geometry recovery, 3D correspondence, understanding of semantics and dynamic changes, and cross-view consistency.</p><p>Its output is therefore a three-dimensional space that can be explored, rendered, computed upon and reused, beyond a video intended for viewing or interaction.</p>","align":"left","textStyles":{}},{"id":"block_6348e1b8-7c62-46cc-9f86-a0aa1650d631","type":"paragraph","subheading":"MoWorld 3D on Huawei Cloud: Connecting Models and Computing","html":"<p>Tianrun Chen and Huawei Cloud Zhejiang general manager Yao Pengzhi jointly announced the official launch of MoWorld 3D on Huawei Cloud. This takes the model beyond the laboratory into a plug-and-play public tool, opening its capabilities to industry and supporting faster deployment of world model applications.</p>","align":"left","textStyles":{}},{"id":"block_06440761-855d-421e-843c-9021fcbb36a7","type":"image","src":"/uploads/companyNews/1786931559275-d1e4a556-30e9-49da-9769-35ccdd7fe36d.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_4b00cad1-58c9-432b-8b5a-c5277056b938","type":"paragraph","subheading":"Partners Join to Build Future Digital City Infrastructure","html":"<p>The value of a model depends on its use in real scenarios as well as its technical metrics. Yang Ming, assistant to KOKONI&#x27;s chairman, presented solutions across industries, describing how foundation-model capabilities can support 3D photo albums, spatial exploration, interior design, street scenes and visualization. The presentation outlined the company&#x27;s route from technical capabilities to industrial value.</p>","align":"left","textStyles":{}},{"id":"block_b3d08689-8425-4ae4-b8ca-9a7a560d3352","type":"image","src":"/uploads/companyNews/1786931578760-423f58c5-ea73-4984-aee5-89d66b7fafaf.jpg","alt":"Yang Ming, assistant to KOKONI's chairman","caption":"","layout":"normal"},{"id":"block_070f99a1-8cd3-45fd-908a-14b8139bd9ea","type":"paragraph","subheading":"","html":"<p>As embodied intelligence enters real environments, its perception and interaction extend from individual local scenes to large, complex urban systems combining production, daily life and public administration. Major cities are becoming vital data sources and testing grounds in the physical AI era.</p><p>Zhou Shangchen, chair of Hangzhou High-tech Digital Commerce Technology Group, introduced the joint development plan. She said future digital city infrastructure would use KOKONI&#x27;s world model capabilities to lower the barriers to urban 3D modeling, enabling government, businesses and the public to contribute spatial data and digital content to a continuously updated 3D urban foundation. This would support high-precision city mapping, dynamic simulation and multiple applications, with full-lifecycle data management to safeguard data security.</p>","align":"left","textStyles":{}},{"id":"block_813e78ec-77d0-48b2-a704-ed3ffbf79d52","type":"image","src":"/uploads/companyNews/1786931632867-733e9186-c790-4dbd-ad2a-11084e8f8448.png","alt":"Zhou Shangchen, chair of Hangzhou High-tech Digital Commerce Technology Group","caption":"","layout":"normal"},{"id":"block_93e2e9b8-311e-4d85-b3ec-4b4375b7e73a","type":"paragraph","subheading":"","html":"<p>Qian Xinyuan, a Huawei Ascend chip product management expert, presented “Ascend Cloud 950: Redefining Architectural Boundaries and a New Paradigm for High-performance Computing.” He explored how the latest Ascend chips support model training through architectural changes and advances in computing performance.</p>","align":"left","textStyles":{}},{"id":"block_a1d6cf56-acbe-4a1a-a105-4419ad607d4b","type":"image","src":"/uploads/companyNews/1786931678533-2328e1da-bf94-4016-8f65-4b331a63ba65.png","alt":"Qian Xinyuan, Huawei Ascend chip product management expert","caption":"","layout":"normal"},{"id":"block_08f09e78-865f-4729-a234-924caa0a85da","type":"paragraph","subheading":"","html":"<p>Representatives from the high-tech division of Hangzhou&#x27;s Development and Reform Commission, Binjiang&#x27;s Development and Reform Bureau, the pilot base, High-tech Digital Commerce Technology and KOKONI then launched the joint initiative. It will bring MoWorld 3D into more real scenarios and accelerate the integration of 3D world models with city-scale spatial data, digital twins and intelligent applications.</p>","align":"left","textStyles":{}},{"id":"block_a21e77c8-4111-44ca-b05c-101189381129","type":"image","src":"/uploads/companyNews/1786931752553-8653ea7d-da45-4905-b753-9edc2edec690.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_7619b482-4bde-4534-a049-ef8f3441f2a5","type":"paragraph","subheading":"Roundtable: The Future of the World Model Industry","html":"<p>KOKONI algorithm scientist Ding Chaotao moderated a roundtable on the future of models and data. Participants included Wang Zonghui, director of the pilot base&#x27;s Frontier Embodied Intelligence Applications and Evaluation Laboratory and a doctoral supervisor at Zhejiang University; Yang Yuhang, chief technical expert at Legend Holdings&#x27; Institute of Advanced Technology; Zhang Xiaobo, R&amp;D director of the embodied intelligence pilot base company; Huawei Ascend chip product management expert Qian Xinyuan; and Huang Peng, CEO of Hangzhou Zhiji Yige Embodied Technology.</p>","align":"left","textStyles":{}},{"id":"block_9d8c4a0a-2cea-4136-b2b7-8e39f19aa2be","type":"image","src":"/uploads/companyNews/1786931778052-e364e48c-dca4-462d-ac26-c724aecfd2a7.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_c8c104d7-4335-4948-8d59-6f4e12361850","type":"paragraph","subheading":"","html":"<p>The panel discussed cooperation between world models and data, domestic computing support, embodied intelligence applications, industrial implementation and commercialization. University research, foundational chips, industrial platforms and practical applications offered complementary perspectives on the field&#x27;s future.</p>","align":"left","textStyles":{}},{"id":"block_bb535fcc-94ab-4b6a-a2a1-f600adf57587","type":"image","src":"/uploads/companyNews/1786931834701-3d2f2035-4937-415e-adb6-1496bd455800.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_115b5a4a-da34-4384-b2dc-f32399f4639b","type":"paragraph","subheading":"From Model R&D to Real-world Applications","html":"<p>The MoWorld 3D release, its launch on Huawei Cloud and the digital city initiative connect model development, domestic computing support and industry validation. They mark both technical progress for KOKONI&#x27;s 3D world models and a move from laboratory research to cloud deployment, real spaces and complex industrial scenarios.</p><p>KOKONI will continue to work within the domestic computing ecosystem to strengthen world models&#x27; understanding and construction of space, geometry and dynamic environments. Together with government, universities and industry partners, it will advance validation and deployment and explore MoWorld 3D in future digital cities, embodied intelligence, digital twins and spatial content production.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_d7f918fc-ce84-4895-ab88-0c087b765349","date":"2026-07-11","time":"2026.07.11","title":"KOKONI Releases MoWorld, Its First Ultra-high-frame-rate Interactive World Model","summary":"From generating content to simulating a world that keeps changing.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/companyNews/1786931086180-c744e150-8d5f-4453-b6d6-f268d8fa8a81.png","mediaType":"image","sortOrder":4,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_6cc1f80d-0877-4c0c-b4ba-9096b099e471","type":"paragraph","subheading":"","html":"<p>Over the past few years, generative AI&#x27;s most striking capability has been content creation. We recently released MoWorld, our first ultra-high-frame-rate interactive world model, and introduced the Flash World Model concept: high frame rates are essential to keeping world models usable in real scenarios. This marks a move from content generation toward interactive world simulation.</p><p>Without relying on high-end GPUs, MoWorld delivers cinematic image quality and real-time interaction at up to 50 FPS, reducing the cost of bringing world models to large-scale applications.</p>","align":"left","textStyles":{}},{"id":"block_a5c7d28e-23b2-4731-9903-fb1cbcccc016","type":"video","src":"/uploads/companyNews/1786930828104-05c78410-ada0-498e-a224-bc3422d0e48a.mp4","poster":"","caption":"","layout":"normal"},{"id":"block_f64a44db-1ad8-440d-b781-38d2c0525845","type":"paragraph","subheading":"A New Cost Structure: Inference at 30% of GPU Costs","html":"<p>MoWorld uses a full-stack Ascend NPU solution, with costs around one-half to two-thirds those of comparable GPUs. For the same 14B MoE model size, taking hardware procurement, cluster networking and long-term operations into account, inference costs can be as low as 30% of the GPU alternative.</p><p>This delivers greater efficiency at lower cost on a fully domestic technology stack, beyond simple hardware substitution.</p>","align":"left","textStyles":{}},{"id":"block_4e783f4e-2206-4a16-ab47-501d2fe59b02","type":"paragraph","subheading":"Flash World Model: Real-time Interaction at Up to 50 FPS","html":"<p>MoWorld optimizes real-time inference across three coordinated layers: the pipeline, parallel processing and operators.</p><p>At the pipeline level, it removes unnecessary overhead, reuses conditional encodings and reserves memory for core generation.</p><p>Across multiple NPUs, it decouples decoding from backbone generation and distributes computing tasks across cards.</p><p>At the execution level, mixed-precision quantization and efficient attention reduce memory transfers and redundant computation.</p>","align":"left","textStyles":{}},{"id":"block_c8b13fb3-ae66-4e05-b3c9-1e9d2184fdba","type":"image","src":"/uploads/companyNews/1786930943917-7fdb3759-1af2-441c-9137-59faea4faf5d.png","alt":"","caption":"","layout":"normal"},{"id":"block_401ac8b7-a7e4-46fa-bcd3-adebf8936aba","type":"paragraph","subheading":"","html":"<p>Together, these three layers enable a 14B-parameter MoE world model to perform real-time inference at up to 50 FPS on Huawei Ascend NPUs.</p>","align":"left","textStyles":{}},{"id":"block_ad1d691e-5079-4c06-a397-9ef4f05c3cc5","type":"paragraph","subheading":"High Quality at Lower Inference Cost","html":"<p>Low-cost inference and high frame rates do not require sacrificing quality. MoWorld demonstrates industry-leading image quality, stability and spatial consistency.</p><p>It maintains stable spatial structures in large aerial scenes, including cities, highways and deserts. Building layouts, road directions and distant details remain clear. Cross-view geometry stays stable even during dives and rapid ascents, keeping the world continuous and consistent.</p><p>Large-scene stability: indoor libraries, Gothic halls, classical streets, pyramid ruins and shopping-mall corridors retain their structures throughout extended exploration.</p><p>Character consistency: subjects retain their appearance across long sequences without deformation or loss of identity.</p><p>Continuous, controllable viewpoints: full six-degree-of-freedom camera control supports an immersive 3D exploration experience.</p>","align":"left","textStyles":{}},{"id":"block_9f8c1f54-d4dc-4d69-a28b-6483c30916bd","type":"paragraph","subheading":"Exploring More Applications with MoWorld","html":"<p>MoWorld marks a key step from validating model capabilities to building a real-time interactive system. As real-time performance, deployment costs and long-term stability become practical, we are exploring further applications:</p><p>Games and interactive entertainment</p><p>Embodied intelligence and autonomous driving</p><p>Film production</p><p>Digital twins and 3D reconstruction</p>","align":"left","textStyles":{}},{"id":"block_cc2de69f-ec5e-42e8-8689-65e67e5e9a87","type":"video","src":"/uploads/companyNews/1786931059438-476588dd-d5f7-4617-93f7-c4b35228974d.mp4","poster":"","caption":"","layout":"normal"},{"id":"block_ecd33c12-76ff-4d53-a0b2-39a2b71a3732","type":"paragraph","subheading":"A Future Built on a Fully Domestic Technology Pipeline","html":"<p>MoWorld&#x27;s release represents a milestone in KOKONI&#x27;s world model development.</p><p>From data preparation, model training and autoregressive distillation to real-time inference on domestic NPUs, MoWorld establishes an integrated technical pipeline and demonstrates the potential to combine high frame rates, low costs and stable generation over long sequences.</p><p>KOKONI will continue to advance core world model technologies, work with the domestic computing ecosystem, explore AI&#x27;s understanding and simulation of space, motion and the physical world, and bring these capabilities to more real applications.</p><p>From generating content to simulating a world that keeps changing.</p><p>KOKONI continues to move forward.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_0bf39f63-48bc-4c89-a053-5e7882cccd34","date":"2026-06-16","time":"2026.06.16","title":"KOKONI Updates: Deepening Partnerships and Expanding the Ecosystem","summary":"KOKONI joins major industry events, working with partners to expand the technology ecosystem.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/1782703058402-9832972d-a718-46de-8ba6-efbfe9964612.png","mediaType":"image","sortOrder":5,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_1782703058264_ahhb86ri","type":"paragraph","subheading":"","html":"<p>01</p><p>AI for Culture and Tourism</p><p>At the Launch of a National AI Application Pilot Base</p><p>The launch meeting for the national AI application pilot base for culture and tourism was recently held in Hangzhou. Guests included Li Yanyi, a member of the Zhejiang Provincial Party Standing Committee and head of its Publicity Department, and Pan Yunhe, an academician of the Chinese Academy of Engineering, who jointly launched the base.</p><p>Invited as a technology company in spatial intelligence, KOKONI exhibited its independently developed 4D world model technologies, highlighting dynamic scene reconstruction, spatial understanding and real-time interaction.</p><p>During the exhibition, officials and other guests visited KOKONI&#x27;s booth to learn about its R&amp;D and applications in spatial intelligence, 3D reconstruction and world models.</p>","align":"left","textStyles":{}},{"id":"block_1782703235379_7420gcqb","type":"image","src":"/uploads/1782703255878-9960206d-fdc5-4a45-8f12-8626f274e08e.png","alt":"","caption":"","layout":"normal"},{"id":"block_1782703285594_eeluz8go","type":"paragraph","subheading":"","html":"<p>02</p><p>Huawei Cloud AI CITY Ecosystem Partner Meeting</p><p>Sharing KOKONI&#x27;s Work on 4D World Models</p><p>A Huawei Cloud AI CITY ecosystem partner cooperation meeting was recently held in Guizhou under a theme of bringing the cloud ecosystem together for a digital and intelligent future. It focused on the AI CITY ecosystem and urban intelligence, with discussions on AI data foundations and embodied intelligence data collection.</p><p>As an ecosystem partner, KOKONI presented its domestically native 4D dynamic world foundation model. The talk introduced its technical approach, core capabilities and applications, highlighting progress in spatial perception, scene understanding and real-world modeling.</p><p>The meeting strengthened communication and collaboration with Huawei Cloud AI CITY partners and created further opportunities for deployment and ecosystem cooperation.</p>","align":"left","textStyles":{}},{"id":"block_1782703377192_0tfpsw4u","type":"image","src":"/uploads/1782703394297-5d459968-07ce-4823-a4f2-51d7a264c5fd.png","alt":"","caption":"","layout":"normal"},{"id":"block_1782703326123_jumw534y","type":"paragraph","subheading":"","html":"<p>03</p><p>Helping AI Understand the Physical World</p><p>KOKONI at INSPIRE 2026</p><p>KOKONI attended INSPIRE 2026 as a Huawei Cloud startup partner. The conference brought together developers, industry experts and ecosystem partners from around the world to discuss AI and cloud technology trends, industrial innovation and ecosystem development.</p><p>KOKONI showcased work presented as a new domestic benchmark for 4D world models. Built on Huawei Ascend Cloud computing, its independently developed KOKONI-WORLD model supports full-HD 4D scene generation within minutes and real-time interaction. Core capabilities include dynamics decoupling, dynamic 4D reconstruction, persistent spatiotemporal memory and high-fidelity geometric reconstruction, with applications in embodied intelligence, autonomous driving, digital twins, film production and 3D creation. KOKONI held in-depth discussions with Huawei Cloud partners and industry guests, and its 4D world model work attracted broad attention.</p>","align":"left","textStyles":{}},{"id":"block_1782703394987_cmt3mpbw","type":"image","src":"/uploads/1782703408037-cc90fa62-e9a6-4f04-a74b-30c74f25d865.png","alt":"","caption":"","layout":"normal"}],"textStyles":{}},{"id":"news_01","date":"2026-06-12","time":"2026.06.12","title":"Zhejiang Provincial Talent Fund Invests in KOKONI","summary":"Supporting young talent and frontier innovation","summaryHtml":"","body":"abcdefg","bodyHtml":"","media":"/uploads/1782703990585-4f21286e-08cf-4b23-ad8e-291eb150dfc2.png","mediaType":"image","sortOrder":6,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"news_block_fa024f07-4393-4264-8e24-7abe2e779e37","type":"paragraph","subheading":"","html":"<p>Introduction</p>","align":"center","textStyles":{}},{"id":"news_block_f1d4f60a-a190-428f-9a2c-f460a824246e","type":"paragraph","subheading":"","html":"<p>The AI industry is moving beyond similar surface-level applications toward the underlying paradigms and commercialization capabilities of 4D world models. At this point, two Zhejiang state-backed investment platforms have both increased their support for KOKONI. Their investments recognize a young research team&#x27;s technical breakthroughs and commercial potential, while positioning provincial capital for future industries and differentiated strengths.</p><p>Zhejiang Venture Capital focuses on the long-term industrial value of 4D world models, while the Provincial Talent Fund supports high-level talent and the commercialization of research. Their complementary strengths support young researchers working at the frontier. The investment advances technology entrepreneurship and reflects Zhejiang&#x27;s long-term approach to future AI industries and homegrown deep technology.</p><p>The following is translated from “Provincial Talent Fund Invests in KOKONI,” originally published by the Zhejiang Talent Development Group&#x27;s WeChat official account.</p><p>The Zhejiang Provincial Talent Development Venture Capital Fund, referred to here as the Provincial Talent Fund, recently completed its investment in Moxin (Hangzhou) Technology Co., Ltd. (KOKONI). The investment focuses on frontier AI, high-level young talent and disruptive technologies, supporting Zhejiang&#x27;s ambition to become a globally influential center for science, technology and AI innovation.</p><p>KOKONI is a frontier AI company specializing in spatial intelligence and world models. Its core architecture is a purely implicit, large-scale, scalable feed-forward neural network. Unlike mainstream approaches that rely on intermediate representations such as 3D Gaussian Splatting (3DGS), its end-to-end training paradigm enables AI to learn and interpret 3D geometry and physical rules directly. The team has explored scaling laws in spatial intelligence, showing that coordinated expansion of data, models and compute can steadily improve spatial understanding. It has achieved breakthroughs in feed-forward dynamic 4D reconstruction, continuous long-sequence spatial modeling and high-resolution 3D reconstruction. Its 10-billion-parameter 4D foundation model has been applied to visual-language question answering and vision-language-action manipulation, with reconstruction accuracy and dynamic-scene support in the industry&#x27;s leading tier. The company brings together young algorithm scientists from leading universities worldwide. Founder Tianrun Chen is a PhD candidate at Zhejiang University, supervised by Pan Yunhe, a Chinese Academy of Engineering academician and leading AI expert. Born after 2000, Chen has received the National Youth Science and Technology Innovation Award, the Central Committee of the Communist Youth League&#x27;s highest award for technological innovation, appeared on Forbes Asia and Hurun under-30 entrepreneur lists, and published research at leading conferences and journals.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_02","date":"2026-06-03","time":"2026.06.03","title":"KOKONI and China Telecom Ningbo Partner on Industrial Applications of 4D World Models","summary":"Exploring intelligent applications for industrial environments together","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/1782711021561-5d900985-25f5-491c-8dcc-c942f4f83d20.jpg","mediaType":"image","sortOrder":7,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_1782711024230_bayggx3x","type":"paragraph","subheading":"","html":"<p>On May 29, Moxin (Hangzhou) Technology Co., Ltd. (KOKONI) and China Telecom&#x27;s Ningbo branch signed a strategic cooperation agreement. Combining strengths in computing and network infrastructure and 4D world models, they will address industrial digitization and intelligent transformation by integrating local compute resources with frontier spatial AI algorithms, and validate and deploy 4D world models in Ningbo&#x27;s industrial environments.</p>","align":"justify","textStyles":{}},{"id":"block_1782711031504_j66s29hr","type":"image","src":"/uploads/1782711212146-0465d157-5713-4835-b963-7cd17be4e5bb.png","alt":"","caption":"","layout":"normal"},{"id":"block_1782711216997_9p859xyx","type":"paragraph","subheading":"","html":"<p>Working at the frontier of physical AI, KOKONI develops foundational 4D world model services, with research in dynamic 3D reconstruction, continuous spatial understanding and interactive scene modeling. Its independently developed KOKONI World model continues to evolve on a domestic technology architecture. Full-HD real-time interaction, long-term scene memory and dynamic spatial reasoning form a technical foundation for intelligent manufacturing and industrial digitization. It has developed customized solutions for industrial digital twins, production-line simulation, 3D inspection of high-end equipment and spatial perception for industrial automation, supporting dynamic 3D modeling, spatial analysis and virtual simulation.</p><p>China Telecom Ningbo is a core builder and operator of the city&#x27;s digital infrastructure. Its citywide integrated computing and network infrastructure supports industrial-grade compute deployment, local network services and large-scale compute scheduling. Through long-standing work with Ningbo&#x27;s manufacturing clusters—including petrochemicals, equipment, automotive parts and smart appliances—it has developed extensive local implementation experience and ecosystem resources in smart factories, industrial internet platforms, enterprise cloud adoption and industrial data governance.</p><p>The partnership focuses on four areas: industrial digital twins, smart factories, intelligent logistics and advanced 3D inspection. China Telecom Ningbo&#x27;s cloud, network and computing capabilities and KOKONI&#x27;s 4D world models will support customizable, practical and replicable intelligent manufacturing solutions. The partners will advance proof-of-concept validation, pilots and reference projects in Ningbo enterprises, supporting production visualization, intelligent inspection, dynamic production-line scheduling and efficient digital management.</p><p>The partners will align frontier spatial AI with local compute and industrial resources. For KOKONI, the collaboration is an important step into Ningbo&#x27;s manufacturing market. China Telecom Ningbo&#x27;s local services and industrial channels will help accelerate commercial deployment, establishing a cycle of technical iteration, local scenario validation and scaled industrial use, and expanding KOKONI&#x27;s work in the real economy.</p><p>For China Telecom Ningbo, KOKONI World adds advanced spatial AI to local computing and network infrastructure, strengthening dynamic industrial 3D modeling, high-precision spatial simulation and long-sequence scene prediction. It extends the intelligent manufacturing service chain into higher-value applications and supports a transition from traditional digital services toward integrated compute, algorithm and scenario solutions.</p><p>As AI becomes more closely integrated with the real economy, the partnership combines frontier algorithms, local compute infrastructure and industrial environments, providing fresh momentum for the digital transformation of Ningbo&#x27;s manufacturers.</p><p>The partners will continue to refine industrial spatial intelligence solutions for Ningbo&#x27;s needs and advance local demonstration projects and scenario testing. Their aim is to improve manufacturing with advanced technology, support Ningbo&#x27;s development as a national reference city for manufacturing digitization, and deliver lasting value from spatial intelligence in physical industries.</p>","align":"justify","textStyles":{}}],"textStyles":{}},{"id":"news_03","date":"2026-05-27","time":"2026.05.27","title":"KOKONI Updates: Advancing Research and Building for the Long Term","summary":"Take a look at our latest updates.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/1782711449193-dcd06387-cc18-4835-86ff-d5b752c133cd.png","mediaType":"image","sortOrder":8,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_1782711451151_t0cpy5k7","type":"paragraph","subheading":"","html":"<p>01</p><p>Innovation at Future Sci-Tech City</p><p>KOKONI Exhibits at a Global AI Industry Conference</p>","align":"center","textStyles":{}},{"id":"block_1782711470327_ftvnbd52","type":"paragraph","subheading":"","html":"<p>The 2026 Global Artificial Intelligence Technology Conference (GAITC), organized by the Chinese Association for Artificial Intelligence, took place at Hangzhou Future Sci-Tech City on May 23–24. More than 200 academicians, experts and industry participants from China and abroad gathered to discuss industry trends.</p><p>The opening ceremony was chaired by Liu Ying, a member of the Hangzhou Municipal Party Standing Committee and Yuhang District Party Secretary. Addresses were delivered by Liu Fei, a member of the Zhejiang Provincial Party Standing Committee and Hangzhou Party Secretary; Weng Qiwen, deputy Party secretary and director of the Ministry of Industry and Information Technology&#x27;s China Centre for Promotion of SME Development; and Dai Qionghai, president of the Chinese Association for Artificial Intelligence and a Chinese Academy of Engineering academician. Academicians Pan Yunhe, Yu Haibin and other experts also attended.</p><p>KOKONI was invited as a representative technology company and showcased its independently developed 4D world model capabilities. During the exhibition, a delegation of ministry officials, academicians and experts visited KOKONI and other key companies to hear about technology development and practical deployment.</p>","align":"left","textStyles":{}},{"id":"block_1782711490538_uv3cy46h","type":"image","src":"/uploads/1782711548880-a5339bc9-940c-4c34-8011-cc1de2589152.jpg","alt":"","caption":"","layout":"normal"},{"id":"block_1782711495750_nw9tw6to","type":"paragraph","subheading":"","html":"<p>02</p><p>Kunpeng &amp; Ascend Developer Conference</p><p>KOKONI Presents Its Research and Receives an Innovation Award</p>","align":"center","textStyles":{}},{"id":"block_1782711498943_om11lwno","type":"paragraph","subheading":"","html":"<p>Huawei hosted the Kunpeng &amp; Ascend Developer Conference (KADC 2026) at Beijing&#x27;s Zhongguancun International Innovation Center on May 22–23. The event brought together technical and research teams to discuss intelligent computing, ecosystem development, technology applications and industry directions.</p><p>KOKONI founder and CEO Tianrun Chen delivered an invited keynote, “Diffusion-based Interactive World Models,” presenting the company&#x27;s frontier work in 4D world models and attracting broad industry attention.</p><p>KOKONI received the Kunpeng &amp; Ascend Dream Launch Innovation Award in recognition of its technological innovation, ecosystem contributions and industry impact.</p>","align":"justify","textStyles":{}},{"id":"block_1782711503606_dk15f3ar","type":"image","src":"/uploads/1782711604353-46804293-8f65-44af-9250-9ead63ca5466.png","alt":"","caption":"","layout":"normal"},{"id":"block_1782711506305_22j7tlg8","type":"paragraph","subheading":"","html":"<p>03</p><p>Named Among Hangzhou&#x27;s Emerging Unicorns</p>","align":"center","textStyles":{}},{"id":"block_1782711509030_i4idt2z0","type":"paragraph","subheading":"","html":"<p>KOKONI was recently named to the 2026 Hangzhou Emerging Unicorn list in the artificial intelligence and embodied intelligence sector, receiving recognition from the city&#x27;s innovation ecosystem and established industry platforms.</p><p>The list highlights high-growth deep-technology companies with strong technical barriers, reflecting professional recognition of their innovation and development potential by industry and investors.</p><p>KOKONI&#x27;s inclusion recognizes its leading 4D world model technology and rapid growth, acknowledging both its technical strengths and future potential.</p>","align":"left","textStyles":{}},{"id":"block_1782711509856_oirepl0s","type":"image","src":"/uploads/1782711647963-f8fb0e0c-78a9-485a-be7c-5e25e8d693c7.png","alt":"","caption":"","layout":"normal"},{"id":"block_1782711653204_ivdbx8t2","type":"paragraph","subheading":"","html":"<p>KOKONI has recently made progress in industry exchange, ecosystem cooperation and company growth. We will continue advancing technology and industrial deployment, building steadily in the field of 4D world models.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_04","date":"2026-05-15","time":"2026.05.15","title":"KOKONI Raises Nearly RMB 100 Million to Strengthen the Foundation for 4D World Models","summary":"KOKONI has completed a new financing round totaling nearly RMB 100 million, jointly backed by Fullhan Microelectronics, Lianrong Zhidao (a Legend Holdings-affiliated fund) and Zhejiang Venture Capital, with follow-on investment from existing shareholders. The participation of domestic chip, smart-device and state-backed industrial capital within the same period reflects growing industry alignment around 4D world models.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/1782725615527-d71037da-848a-4685-a10c-cb8372f034fb.jpg","mediaType":"image","sortOrder":9,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_1782725640262_bv21vlci","type":"paragraph","subheading":"Three Core Technologies Help AI Understand the 3D Physical World","html":"<p>AI&#x27;s capabilities are rapidly expanding from text, images and video into the three-dimensional physical world. In embodied intelligence, autonomous driving, digital twins and industrial simulation, AI must understand where objects are, how spaces change and how scenes remain consistent over time, as well as what appears in an image. KOKONI has focused on these questions since its founding.</p><p>Through training and validation on large-scale data, KOKONI&#x27;s world models have achieved systematic breakthroughs in feed-forward dynamic 4D scene reconstruction, continuous long-sequence spatial modeling and high-resolution feed-forward 3D reconstruction. The team has established an approach to training feed-forward 4D reconstruction and generation networks at scale. Its 4D foundation model has also demonstrated value in VLM question answering and VLA manipulation. A new generation of feed-forward reconstruction models developed with KOKONI&#x27;s participation has entered the industry&#x27;s leading tier in scale (over 10B parameters), reconstruction accuracy and dynamic support, and has earned recognition from customers in consumer entertainment and embodied intelligence.</p>","align":"left","textStyles":{}},{"id":"block_1782725698151_8yaworzt","type":"paragraph","subheading":"Four Industry Pillars Converge","html":"<p>One of the round&#x27;s clearest signals is the arrival of investors from several parts of the industrial chain within a short period.</p><p>Fullhan Microelectronics, a leading listed Chinese company in intelligent vision processing chips, has invested strategically in KOKONI and established close cooperation in dynamic 4D world models. Lianrong Zhidao participated through its affiliate Hanlian Zhixin, bringing a global perspective on smart devices and computing; it will work with KOKONI on applications including embodied intelligence. Zhejiang Venture Capital, a provincial state-backed technology investment platform, brings its industrial investment approach and experience across the deep-technology value chain.</p><p>Together with Huawei&#x27;s Hubble Investment, which invested a month earlier, these shareholders span four industrial pillars: computing infrastructure, core chips, smart devices and state-backed industrial capital. It is unusual for major industrial participants with distinct ecosystems to invest in the same startup at the same time. As AI moves from generating text and images toward understanding and interacting with the 3D physical world, the entire chain—from chips to devices—needs shared foundational spatial understanding. KOKONI&#x27;s accumulated expertise and frontier results have made it a common choice for these investors.</p>","align":"left","textStyles":{}},{"id":"block_1782725752665_2v0tnqon","type":"paragraph","subheading":"Accelerating Development and Opening the Ecosystem","html":"<p>Founder and CEO Dr. Tianrun Chen said the financing will primarily support the evolution of general-purpose 4D world models, accelerate KOKONI-World iteration and public testing, expand multimodal data collection and training pipelines, and open the ecosystem to partners in embodied intelligence, consumer entertainment and digital twins.</p>","align":"left","textStyles":{}}],"textStyles":{}},{"id":"news_6430bbbf-7204-40c1-83b1-a0e9da9db0dc","date":"2026-05-15","time":"2026.05.15","title":"Reconstructing 3D Worlds from Dynamic Video: VGGT-Ω Sets a New Standard for Feed-forward 4D Modeling","summary":"VGGT-Ω, developed with participation from KOKONI researchers, advances feed-forward 3D reconstruction into dynamic scenes, complex viewpoints and long video sequences. It predicts camera parameters and depth directly from multiple images and delivers more stable reconstruction on both static and dynamic benchmarks.\n\nFor spatial intelligence and world models, reconstruction is a foundation for understanding real space, beyond recovering geometry. Real environments consist of continuous movement, viewpoint changes and long sequences rather than still images. VGGT-Ω shows how feed-forward reconstruction is moving toward video inputs closer to the real world.","summaryHtml":"","body":"","bodyHtml":"","media":"/uploads/1782725973165-4f698ffa-8bec-44e1-b0b5-dbb74f4ef5b1.png","mediaType":"image","sortOrder":10,"isActive":true,"slug":"","detailTitle":"","detailBlocks":[{"id":"block_1782725998399_x2011tdy","type":"paragraph","subheading":"Spatial Reconstruction for Real Dynamic Scenes","html":"<p>Change is normal in the real world. Traffic, moving people, rotating cameras and changing viewpoints create complex spatial relationships for reconstruction. A model must distinguish camera motion from the independent movement of objects within a scene.</p><p>VGGT-Ω demonstrations cover traffic flows, tennis player trajectories, natural landscapes and underwater coral. The model recovers static spatial structures while maintaining spatial consistency in dynamic content.</p><p>Feed-forward reconstruction is therefore beginning to handle realistic video inputs, beyond ideal conditions involving a camera moving around a static object. In embodied intelligence, autonomous driving, digital twins and spatial computing, dynamic reconstruction determines whether a model can perceive continuous environments reliably.</p>","align":"left","textStyles":{}},{"id":"block_1782725999877_s56hljnf","type":"paragraph","subheading":"Global Consistency under Complex Motion and Viewpoint Changes","html":"<p>Reconstruction challenges arise from changing viewpoints, repeated textures and sparse structures as well as moving objects. The paper compares VGGT-Ω with MegaSaM, Depth Anything 3 and other methods. Existing approaches can exhibit geometric drift, repeated structures or distorted global layouts in aerial scenes, sparse interiors, repetitive snowy environments and drone footage with substantial camera roll.</p><p>In aerial scenes, MegaSaM can show geometric inconsistencies and texture ghosting that stretch the overall structure. In repetitive scenes such as a snowy cable-car route, Depth Anything 3 can underestimate camera movement. It can also produce duplicate tower structures when a drone circles a tower with large camera roll.</p><p>VGGT-Ω instead produces more complete, continuous spatial structures. It maintains stronger global consistency in sparse, dynamic and aerial scenes and under substantial viewpoint changes, reducing drift, misalignment and duplicated structures.</p>","align":"left","textStyles":{}},{"id":"block_1782726001159_8hy9xyn9","type":"paragraph","subheading":"Significant Improvements on Dynamic Scenes","html":"<p>Quantitative results show leading performance across several static and dynamic benchmarks. The paper evaluates static datasets including 7 Scenes, NRGBD and ETH3D, and dynamic datasets including DyCheck, Sintel and TUM-Dynamic.</p><p>On Sintel, camera-estimation AUC@3° increases from 22.5 to 40.0, and AUC@30° from 58.3 to 79.1. For depth estimation, δ1.25 rises from 74.1 to 93.5. The paper also reports that VGGT-Ω is around 50 times faster than MegaSaM.</p><p>These improvements mean more accurate camera poses, more reliable depth predictions and more stable 3D reconstruction. This is particularly valuable for dynamic video, where spatial understanding must persist through motion, occlusion and changing views.</p>","align":"left","textStyles":{}},{"id":"block_1782726146857_5ysp4ov0","type":"paragraph","subheading":"Efficiency with Long Input Sequences","html":"<p>Practical 3D reconstruction must handle more than short inputs. Robotics, autonomous driving, spatial exploration and digital twins require understanding changes over longer periods.</p><p>VGGT-Ω also demonstrates efficiency advantages. On a single 80 GB A100 GPU, experiments show that VGGT and VGGT-Ω can process more than 1,000 frames, while DA3 reaches memory limits at around 750 frames. Design choices including register attention deliver an approximately 20–25% speed improvement under default settings.</p><p>Efficiency affects whether models can support longer sequences and more frequent video-based spatial understanding. Feed-forward reconstruction provides camera and depth estimates in a single forward pass, without repeated optimization for every scene.</p>","align":"left","textStyles":{}},{"id":"block_1782726148027_ft36buof","type":"paragraph","subheading":"From Reconstruction to Spatial Representations","html":"<p>VGGT-Ω demonstrates both more stable reconstruction and the potential of 3D reconstruction as a training task for spatial intelligence.</p><p>From multiple images, the model learns geometry, camera motion, depth relationships and dynamic changes. These capabilities can support reconstruction itself and potentially robot manipulation, visual-language understanding and scene-level spatial reasoning. The paper notes that scene tokens can act as geometry-aware representations to improve robot VLA tasks.</p><p>Seen this way, 3D reconstruction is evolving from an independent visual geometry task into an important route for world models to understand real space. Future spatial intelligence systems must understand object positions, scale, motion relationships and continuous changes across viewpoints, as well as recognize objects in images.</p><p>VGGT-Ω experiments show feed-forward spatial reconstruction moving closer to the real dynamic world. Continued expansion of models, datasets and dynamic video training can broaden spatial understanding applications in embodied intelligence, autonomous driving, digital twins and 3D content generation.</p><p>Paper: https://arxiv.org/pdf/2605.15195</p>","align":"left","textStyles":{}}],"textStyles":{}}],"companyIntroVideo":{"media":"/uploads/companyIntroVideo/1782821304102-61465db5-6a45-46dc-9d12-0316996339ef.mp4","mediaType":"video","poster":"","title":"KOKONI 4D World Model Video","description":"KOKONI, aiming to become the infrastructure builder for 4D world models. We are developing a high-general, high-precision, interactive 4D dynamic scene foundation model which born for real-time reconstruction, intelligent editing, and novel-view synthesis from 3D to 4D. Enabing precise mapping and efficient simulation of the physical world, thereby profoundly empowering fields such as autonomous driving, embodied intelligence, digital twins, film production, and creative content generation.","descriptionHtml":"","textStyles":{"title":{"fontSize":16}}},"products":[{"id":"world-model-terminal","name":"KOKONI AI 3D Printer","oneLiner":"Self-developed KOKONI AI system, facilitating modeling and creation with zero barriers to entry.","oneLinerHtml":"","learnMoreUrl":"https://kokoni.ltd/home/","description":"The independently developed AI algorithm employs deep learning for image feature extraction, performs depth estimation and model construction within three-dimensional space, and is capable of generating a full three-dimensional high-fidelity model solely from images.","descriptionHtml":"","tags":[],"scenarios":[],"media":"/uploads/products/1782875421209-cc4f59d6-0243-401d-9c23-0f9ee9ec7508.mp4","mediaType":"video","sortOrder":1,"isActive":true,"textStyles":{}},{"id":"kokoni-all-in-one","name":"KOKONI 3D Reconstruction All-in-One Machine","oneLiner":"A Domestic One-Stop Solution for 3DGS/Point Cloud Reconstruction","oneLinerHtml":"","learnMoreUrl":"","description":"Empowering a Myriad of Industries and Constructing a Digital Twin Universe","descriptionHtml":"","tags":["Scenic Area Tours","Terrain Visualization","Factory Digital Twins"],"scenarios":["Scenic Area Tours","Terrain Visualization","Factory Digital Twins"],"media":"/uploads/1782710101769-5c9e23c8-a3bf-4e09-8d21-999acb3b1168.mp4","mediaType":"video","sortOrder":2,"isActive":true,"textStyles":{}},{"id":"data-capture-device","name":"KOKONI EGO Data Acquisition System","oneLiner":"Builds a new \"human-machine collaborative\" smart factory operation mode","oneLinerHtml":"","learnMoreUrl":"","description":"Seamlessly integrate smart wearable devices, SCADA (Supervisory Control and Data Acquisition) and MES (Manufacturing Execution System) to create a real-time data closed loop from management to execution levels, empower frontline employees, and achieve comprehensive digitization, intelligence, and efficient collaboration in the production process.","descriptionHtml":"","tags":[],"scenarios":[],"media":"/uploads/1782718022318-7557144a-b3a6-41d6-afec-7c3a46aefade.png","mediaType":"image","sortOrder":3,"isActive":true,"textStyles":{}}],"jobs":[{"id":"developer-growth-operator","title":"Brand Specialist","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Marketing","cities":["Hangzhou"],"tags":[{"label":"New","color":"blue"},{"label":"Urgent","color":"red"}],"description":"Brand communications","descriptionHtml":"","responsibilities":["1. Maintain and update the company's WeChat official account, website and other platforms; write promotional materials and brand presentations; and handle editing, review, proofreading and basic channel operations.","2. Research markets, competitors and user needs, analyze industry trends, produce data reports and identify opportunities to support brand strategy.","3. Help organize and run press conferences, industry forums and competitions, company culture activities and other brand events.","4. Commission contributions from internal and external contacts and collect, organize and archive promotional assets and coverage.","5. Plan, edit and operate short-form video content to support the brand across multiple platforms.","6. Complete other tasks assigned by management."],"responsibilitiesHtml":"","requirements":["1. A background in journalism, advertising, communications, digital media, Chinese language and literature or a related field; AI or technology industry experience is preferred.","2. At least one year of experience in social media operations, content editing or copywriting, with strong writing skills; proficiency in WeChat layout tools such as Xiumi and 135, and basic skills in Photoshop, Illustrator and Jianying/CapCut.","3. A strong ability to learn, attention to detail, effective cross-team collaboration, and resilience under pressure."],"requirementsHtml":"","createdAt":"2026-01-01T00:07:00.000Z","firstPublishedAt":"2026-01-01T00:07:00.000Z","updatedAt":"2026-07-01T06:01:32.059Z","isActive":true,"textStyles":{}},{"id":"ai-talent-acquisition-lead-partner","title":"Finance Specialist","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Corporate","cities":["Hangzhou"],"tags":[{"label":"Urgent","color":"red"},{"label":"New","color":"orange"}],"description":"Finance and accounting","descriptionHtml":"","responsibilities":["1. Accounting and bookkeeping: handle daily expense reimbursements, procurement payments, revenue recognition and related entries accurately, promptly and in compliance with requirements.","2. Tax filing and management: complete VAT, individual income tax, corporate income tax and other filings on time, and support annual tax inspections and reconciliations.","3. Financial reporting and analysis: assist with monthly and quarterly statements, expense analysis and budget tracking, and provide financial data to business teams.","4. Cash and asset management: perform bank reconciliations, coordinate cash allocation and count fixed assets to safeguard company funds and assets.","5. Internal controls and compliance: support annual audits, financial inspections and contract reviews, and help improve financial policies and procedures."],"responsibilitiesHtml":"","requirements":["1. A bachelor's degree or above in finance, accounting, auditing, economics or a related field.","2. Three to five years of auditing experience at an accounting firm; IPO project experience is preferred.","3. Familiarity with accounting standards and national tax policies, the ability to independently handle a full set of accounts, and proficiency in Kingdee, Yonyou or comparable financial software and Excel.","4. A CPA qualification or an intermediate or higher accounting qualification is an advantage.","5. Attention to detail, clear reasoning, a strong sense of responsibility, effective communication and teamwork, and the ability to understand the business quickly."],"requirementsHtml":"","createdAt":"2026-01-01T00:03:00.000Z","firstPublishedAt":"2026-01-01T00:03:00.000Z","updatedAt":"2026-06-29T03:01:14.270Z","isActive":true,"textStyles":{}},{"id":"product-design-intern","title":"E-commerce Customer Service Representative","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Customer Service","cities":["Huzhou"],"tags":[{"label":"New","color":"blue"},{"label":"Urgent","color":"red"}],"description":"E-commerce customer service","descriptionHtml":"","responsibilities":["1. Respond to presales inquiries on Tmall, JD.com, Douyin and other platforms, answer product questions, guide customers through purchases and improve inquiry conversion rates.","2. Identify customer needs and recommend suitable 3D printers and consumables to increase order value and cross-sales.","3. Stay informed about store promotions, proactively contact interested customers, close transactions and achieve individual sales targets.","4. Handle pricing, negotiation, logistics and other daily presales inquiries, and support the operations team with promotion preparation and customer outreach.","5. Record common questions and customer feedback, help improve response scripts, and regularly report presales conversion data."],"responsibilitiesHtml":"","requirements":["1. A background in e-commerce, marketing, business administration or a related field is preferred.","2. At least one year of e-commerce customer service experience; presales experience on Tmall, JD.com or Douyin is preferred.","3. A typing speed of more than 60 Chinese characters per minute.","4. Willingness to work shifts: 08:30–17:30 or 15:30–23:00.","5. Strong communication skills, a sales mindset, resilience under pressure and a clear focus on performance."],"requirementsHtml":"","createdAt":"2026-01-01T00:02:00.000Z","firstPublishedAt":"2026-01-01T00:02:00.000Z","updatedAt":"2026-06-29T02:59:57.469Z","isActive":true,"textStyles":{}},{"id":"10x-team-researcher-intern","title":"Mid-level Compute Operations Engineer","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Operations","cities":["Hangzhou"],"tags":[{"label":"Priority Role","color":"purple"},{"label":"Urgent","color":"red"}],"description":"Compute resource operations","descriptionHtml":"","responsibilities":["1. Internal compute controls: maintain and reconcile compute asset records, review IDC bills and associated risks, manage service-provider contracts throughout their lifecycle, and reconcile and settle charges.","2. External supplier management: coordinate with purchased-compute suppliers, manage the full resource lifecycle, assess supplier performance, and resolve SLA disputes.","3. Government and enterprise customer operations: match compute demand with resources, develop customized delivery plans, manage contracts, invoicing and collections, coordinate delivery and operational issues, and support renewals and new opportunities.","4. Industry and technology analysis: understand GPU servers, InfiniBand networks and distributed storage infrastructure; track NVIDIA chip developments and bare-metal compute pricing; and participate in procurement selection and negotiations."],"responsibilitiesHtml":"","requirements":["A bachelor's degree or above in science or engineering, with at least three years of experience in compute, cloud or IDC operations, procurement or commercial management.","Familiarity with enterprise contract approval and signing procedures, with end-to-end experience handling project contracts.","An understanding of basic GPU computing terminology and the ability to coordinate with technical teams.","Strong spreadsheet and resource-record management skills, communication and collaboration skills, and the ability to manage concurrent activities.","Preferred qualifications:","Experience with large AI models, cloud compute providers or GPU procurement.","Experience in compute resource resale or building supplier SLA/KPI management systems."],"requirementsHtml":"","createdAt":"2026-01-01T00:01:00.000Z","firstPublishedAt":"2026-01-01T00:01:00.000Z","updatedAt":"2026-06-29T02:58:26.202Z","isActive":true,"textStyles":{}},{"id":"10x-team-fellowship","title":"Mechanical Design Engineer","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Design","cities":["Huzhou"],"tags":[{"label":"Urgent","color":"red"}],"description":"Product mechanical and structural design","descriptionHtml":"","responsibilities":["1. A bachelor's degree or above in mechanical design, materials engineering or a related field is preferred. An associate degree may be accepted for candidates with more than five years of relevant industry experience.","2. At least three years of experience in complete mechanical design for electronic products; exceptional candidates with two years of experience may be considered.","3. Clear communication, a professional attitude and no harmful habits."],"responsibilitiesHtml":"","requirements":["1. Proficient in mainstream design tools such as Creo, SolidWorks and AutoCAD, with a solid foundation in exterior and surface modeling.","2. Strong mechanical analysis skills and familiarity with the properties of common metallic and nonmetallic materials.","3. Proficient in essential office software, including Excel, Word and PowerPoint.","4. Good English reading and writing skills."],"requirementsHtml":"","createdAt":"2026-01-01T00:00:00.000Z","firstPublishedAt":"2026-01-01T00:00:00.000Z","updatedAt":"2026-06-29T02:57:16.503Z","isActive":true,"textStyles":{}},{"id":"3d-generation-algorithm-intern","title":"Software Development Intern","recruitmentType":"campus","recruitmentTypeLabel":"Internships","category":"Research & Development","cities":["Hangzhou"],"tags":[{"label":"New","color":"blue"},{"label":"Internship","color":"orange"}],"description":"Software development","descriptionHtml":"","responsibilities":["1. Develop software, APIs and features for internal projects, platform systems and business applications.","2. Build Python scripts, automation tools, data processing utilities and tools that call model services.","3. Deploy project environments, release and monitor services, investigate logs, locate faults and resolve issues.","4. Support project managers with customer-site deployment, testing, demonstrations and system delivery.","5. Write essential technical documentation, including API references, deployment guides, operations manuals and testing notes.","6. Help turn AI model services, MaaS platforms, data processing platforms and intelligent application systems into products according to project needs."],"responsibilitiesHtml":"","requirements":["1. Proficiency in Python for data processing, API development, automation scripts and service integration.","2. Familiarity with common Linux commands and the ability to deploy services, analyze logs and troubleshoot independently.","3. Familiarity with Docker, Nginx, Git, MySQL/PostgreSQL/Redis and other common development and operations tools.","4. Knowledge of AI model inference services, model API integration, large model applications, vector databases or RAG is preferred.","5. Development experience with cloud platforms, government or enterprise projects, tender-related projects, MaaS platforms, large model service platforms or data governance platforms is preferred.","6. Good coding standards, documentation habits and teamwork, with the ability to meet project development and delivery milestones."],"requirementsHtml":"","createdAt":"2026-01-01T00:09:00.000Z","firstPublishedAt":"2026-01-01T00:09:00.000Z","updatedAt":"2026-06-29T03:13:18.867Z","isActive":true,"textStyles":{}},{"id":"multimodal-world-model-researcher","title":"Large Model Algorithm Intern","recruitmentType":"campus","recruitmentTypeLabel":"Internships","category":"Algorithms","cities":["Hangzhou"],"tags":[{"label":"Now Hiring","color":"orange"},{"label":"Priority Role","color":"purple"},{"label":"Remote","color":"orange"},{"label":"Potential Full-time Offer","color":"orange"}],"description":"Research video world models, multimodal understanding and spatial representation learning, and bring frontier model capabilities into product scenarios where they can be evaluated.","descriptionHtml":"","responsibilities":["1. Algorithm R&D support: contribute to module development, training optimization and experiment analysis for spatial intelligence models with 1B–100B parameters.","2. Multimodal data modeling: help process and align 3D vision, language, actions and other modalities.","3. 3D generative model research: implement and optimize scene generation and editing with diffusion models, NeRF, 3DGS and related techniques.","4. Engineering collaboration: work with product and engineering teams to deploy and iterate algorithms in real business scenarios.","5. Research tracking: follow new papers from CVPR, ICCV, NeurIPS and other leading conferences, and contribute to technology reviews and experiment reproduction.","6. Academic output: with support from mentors and the team, develop research for submission to leading international conferences or journals."],"responsibilitiesHtml":"","requirements":["1. A bachelor's degree or above in computer science, artificial intelligence, automation, mathematics or a related field; master's and PhD candidates are preferred.","2. A solid foundation in machine learning and deep learning, with familiarity with Transformer, Diffusion, NeRF/3DGS and other mainstream architectures.","3. Proficiency in Python and familiarity with PyTorch, supported by practical project or research experience.","4. Strong self-motivation, willingness to take on challenges, and a keen interest in spatial intelligence and large multimodal models.","Preferred qualifications:","Experience publishing or submitting papers to leading conferences such as CVPR, ICCV, ECCV, NeurIPS or SIGGRAPH.","Experience training, fine-tuning or deploying models with at least 1B parameters.","Project experience with NeRF, 3DGS, 3D diffusion models or related techniques.","Contributions to open-source projects or technical blogs."],"requirementsHtml":"","createdAt":"2026-01-01T00:08:00.000Z","firstPublishedAt":"2026-01-01T00:08:00.000Z","updatedAt":"2026-06-29T03:10:11.336Z","isActive":true,"textStyles":{}},{"id":"world-model-product-manager","title":"Large Model Algorithm Engineer","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Algorithms","cities":["Hangzhou"],"tags":[{"label":"Now Hiring","color":"orange"},{"label":"Priority Role","color":"orange"}],"description":"Algorithms for spatial intelligence and world models","descriptionHtml":"","responsibilities":["1. Algorithm design and implementation: contribute to system design, core module development and training strategy optimization for spatial intelligence models with 1B–100B parameters.","2. Multimodal understanding and fusion: research and implement efficient alignment and joint modeling of 3D vision, language, actions and other modalities.","3. 3D generation and editing: develop scene generation, editing and controllable synthesis using diffusion models, NeRF, 3DGS and related techniques.","4. Implementation and collaboration: work with product and engineering teams to apply and improve algorithms in real business scenarios.","5. Frontier research: track new developments, contribute to technology reviews and papers, and help build the team's technical influence."],"responsibilitiesHtml":"","requirements":["1. A master's degree or above in computer science, artificial intelligence, robotics or a related field; a PhD is preferred.","2. A solid foundation in deep learning and 3D vision, with familiarity with Transformer, Diffusion, NeRF/3DGS and other mainstream model architectures.","3. Proficiency in Python and PyTorch, with experience training and debugging large-scale models.","4. Strong self-motivation, excellent communication and enthusiasm for defining the next generation of spatial intelligence technology.","Preferred qualifications:","Relevant publications at leading conferences or journals such as CVPR, ICCV, ECCV, NeurIPS or SIGGRAPH.","Experience training, tuning or deploying vision, language or multimodal models with at least 1B parameters.","In-depth research and project results in NeRF, 3DGS, 3D diffusion models or dynamic scene modeling."],"requirementsHtml":"","createdAt":"2026-01-01T00:06:00.000Z","firstPublishedAt":"2026-01-01T00:06:00.000Z","updatedAt":"2026-06-29T03:05:42.940Z","isActive":true,"textStyles":{}},{"id":"10x-team-researcher","title":"Data Collection Engineer","recruitmentType":"campus","recruitmentTypeLabel":"Internships","category":"Data Collection","cities":["Guiyang"],"tags":[{"label":"Now Hiring","color":"orange"},{"label":"Potential Full-time Offer","color":"orange"},{"label":"Remote","color":"orange"}],"description":"3D data collection","descriptionHtml":"","responsibilities":["1. An associate degree or above in any discipline; computer science or data-related fields are preferred.","2. Proficiency in office software. Strong Excel skills and some Python or data analysis skills are an advantage.","3. Attention to detail, good communication, a sense of responsibility and a proactive approach to solving problems."],"responsibilitiesHtml":"","requirements":["1. Carry out daily data collection, quality checks, cleaning, organization and standardized storage, and maintain statistics.","2. Help test new data management tools and work with the R&D team to improve collection and quality-control tools.","3. Complete other tasks assigned by your supervisor."],"requirementsHtml":"","createdAt":"2026-01-01T00:05:00.000Z","firstPublishedAt":"2026-01-01T00:05:00.000Z","updatedAt":"2026-06-29T03:04:06.057Z","isActive":true,"textStyles":{}},{"id":"legal-bp-overseas","title":"Sales Assistant","recruitmentType":"social","recruitmentTypeLabel":"Experienced Hiring","category":"Assistant","cities":["Hangzhou"],"tags":[],"description":"Sales support","descriptionHtml":"","responsibilities":["1. Support the sales team throughout the commercial contract lifecycle, including drafting, following up on reviews, signing and archiving.","2. Track and manage invoice requests, issuance and delivery from start to finish.","3. Track and reconcile customer payments and company receipts to keep records clear and accurate.","4. Compile sales performance data regularly and assist with related reports.","5. Maintain customer files and contract registers, keeping information complete and current.","6. Assist with daily sales administration and other tasks assigned by management."],"responsibilitiesHtml":"","requirements":["1. An associate degree or above; finance, administrative management, business administration or related fields are preferred.","2. Approximately one year since graduation, with around one year of experience in commercial support or contract administration.","3. Proficiency in Word, Excel and PowerPoint, with strong data-handling skills.","4. A careful, responsible approach to work, good communication and coordination skills, and a collaborative mindset.","5. Strong time management skills and the ability to handle several tasks efficiently.","An outgoing personality, an organized approach and a strong service mindset."],"requirementsHtml":"","createdAt":"2026-01-01T00:04:00.000Z","firstPublishedAt":"2026-01-01T00:04:00.000Z","updatedAt":"2026-06-29T03:02:32.564Z","isActive":true,"textStyles":{}}]}