AnswersPublished17 sources
How Hesai’s Kosmo Turns Real Spaces Into Reusable 3D Worlds
Hesai launched Kosmo on October 14, 2026: its Kosmo One camera captures spaces at up to 8K, while exported files reportedly top out at 6K. Hesai says a typical 100 square meter indoor space takes about 15 minutes to scan; Kosmo World can generate 3D Gaussian Splatting, mesh and depth data from one reconstruction.[32...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Hesai launched Kosmo on October 14, 2026: its Kosmo One camera captures spaces at up to 8K, while exported files reportedly top out at 6K.
- Hesai says a typical 100 square meter indoor space takes about 15 minutes to scan; Kosmo World can generate 3D Gaussian Splatting, mesh and depth data from one reconstruction.[32]
- In a Dairy Queen example, Hesai reported a 99% task success rate for a Sharpa robot after scanning and reconstructing the store; the supplied reporting does not establish this as an independently validated result.[4]
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is Kosmo, the 3D world data engine launched by automotive lidar maker Hesai on October 16, 2026, and how does its Kosmo One camera—clai. Article summary: Kosmo is Hesai’s camera-and-cloud system for converting real environments into reusable 3D digital spaces for robot training, world-model development and immersive content creation.[7][13] One correction: Hesai’s own lau. Topic tags: general, general web, government. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake n
Kosmo is Hesai’s camera-and-cloud system for turning real environments into reusable 3D digital spaces. The lidar maker launched it on October 14, 2026, positioning the platform for robot training, world-model development and immersive content creation.23 Its core idea is to capture a physical space once, then process that scan into different kinds of 3D data.
How Kosmo One captures a space
Kosmo One combines high-resolution lidar, an 8K imaging system built around a one-inch sensor, and a high-precision inertial measurement unit (IMU). Hesai says the system synchronizes capture of a space’s geometry, color, texture and device pose at microsecond precision.32
The camera’s measurements are processed using 3D Gaussian Splatting (3DGS) and image-processing software. Hesai claims millimeter-level geometric accuracy and high-fidelity reproduction of materials, colors, textures and light and shadow. Those are manufacturer claims; the available sources do not provide independent test results confirming the accuracy specification.32
Hesai says a typical 100-square-meter indoor space can be scanned in about 15 minutes. Its launch materials also describe features such as scanning guidance and the ability to resume a scan, intended to simplify capture.32 The advertised 8K figure needs a qualification: reporting says exported files are capped at 6K, so 8K capture does not mean every output is available in 8K.
1
2
What Kosmo World does with the scan
Kosmo World is the cloud platform for creating and managing 3D content. Its underlying Kosmo Spatial Core is described as producing 3DGS, mesh and depth data from a single reconstruction.32 These formats give a scan different forms for visual presentation, geometry and depth information; the available sources do not establish a guarantee that all outputs will be identical in every detail.
Hesai also describes generative features that fill in unscanned regions, reconstruct materials and textures, and combine information from multiple views into a navigable scene.32 A filled-in area is an AI-generated completion, not a direct measurement of a place the camera scanned. That distinction matters when judging a reconstructed scene for tasks where exact physical detail is important.
The platform is aimed at reusing captured spaces for robotics and other spatial workflows. Reporting says Kosmo can generate more than 10,000 robot-training trajectories from one reconstructed space.1 That figure describes the company’s reported capability; it does not, on its own, show how useful or representative those generated trajectories are for a particular robot or task.
What Hesai reported from a Dairy Queen deployment
Hesai reported that scanning and reconstructing a Dairy Queen store took about 15 minutes, after which a Sharpa robot operated there with a 99% task-success rate.4 Other coverage describes the comparison as roughly 300 hours of manual environment creation versus 15 minutes using Kosmo.
2
8
These figures are an example reported by Hesai, not a controlled comparison verified by the sources available here. The supplied reporting does not detail the tasks behind the success rate or establish that the 15-minute figure includes robot training and deployment time. Treat the numbers as a company-reported case study rather than a general performance guarantee.
Beyond robotics: partnerships and creative applications
At launch, Hesai announced partnerships with embodied-AI companies Galbot and Dexmal, and showcased applications developed with Pixar’s RenderMan and Amap.17 That suggests the company is presenting Kosmo as a platform for both robotics and visual-content or mapping workflows. The launch materials do not, by themselves, demonstrate the performance of every potential integration or application.
What to take away
Kosmo’s central proposition is a capture-to-reconstruction pipeline: Kosmo One records a physical environment, and Kosmo World turns the data into 3D assets that can be reused across training and content workflows.23
32 Its practical appeal will depend on more than capture resolution: export limits, reconstruction quality, the treatment of unscanned areas and the usefulness of generated training data all matter. Hesai’s launch figures offer an initial picture of the system, but its accuracy and deployment claims should be read as company-reported until supported by independent evaluation.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "How Hesai’s Kosmo Turns Real Spaces Into Reusable 3D Worlds"?
Hesai launched Kosmo on October 14, 2026: its Kosmo One camera captures spaces at up to 8K, while exported files reportedly top out at 6K.
What are the key points to validate first?
Hesai launched Kosmo on October 14, 2026: its Kosmo One camera captures spaces at up to 8K, while exported files reportedly top out at 6K. Hesai says a typical 100 square meter indoor space takes about 15 minutes to scan; Kosmo World can generate 3D Gaussian Splatting, mesh and depth data from one reconstruction.[32]
What should I do next in practice?
In a Dairy Queen example, Hesai reported a 99% task success rate for a Sharpa robot after scanning and reconstructing the store; the supplied reporting does not establish this as an independently validated result.[4]