World models are becoming a foundation for physical AI. By creating realtime, interactive environments, they can change how robots learn, train, and adapt before entering the real world. Great to see Decart featured in Inc. Magazine on this shift. Read the article >> https://lnkd.in/dh4PZDyf
Decart
Technology, Information and Internet
San Francisco, California 16,154 followers
An efficiency-first AI research lab, building foundational realtime world and video models.
About us
Our efficient systems-level AI infrastructure enables a tenfold improvement in both training and inferencing of the largest generative models. Leveraging these capabilities, Decart is able to train its own foundational generative interactive models and make it accessible for everyone in real-time.
- Website
-
https://decart.ai
External link for Decart
- Industry
- Technology, Information and Internet
- Company size
- 51-200 employees
- Headquarters
- San Francisco, California
- Type
- Privately Held
- Founded
- 2023
Locations
-
Primary
Get directions
San Francisco, California, US
Employees at Decart
Updates
-
A great live conversation with TBPN. Our CEO Dean Leitersdorf joined John Coogan and Jordi Hays to talk about Decart, the shift to realtime AI, and what it takes to build the infrastructure behind the next generation of models. Thanks for having us. Watch the full episode >> https://lnkd.in/dv3mp_YX
-
Excited to be featured in The Wall Street Journal. Robbie Whelan covers Decart’s work on building the next-gen inference and training platform - so AI labs can run across any hardware, faster. Read the full story >> https://lnkd.in/eSS8TWkp
-
-
A big week for Decart. Our $300M Series B marks an important step in scaling Realtime AI and infra, and in bringing frontier world models from research into real products and experiences. Thank you to our team, investors, partners, and everyone supporting us on this journey. We’re just getting started.
-
-
Decart reposted this
Excited to share that we’ve raised $300M in our Series B round, led by Radical Ventures and Jordan Jacobs, bringing our total funding to more than $450M, with leading technology companies joining as both customers and investors. We continue accelerating the path to AGI through our two core pillars: ultra-optimized infrastructure for AI workloads, and realtime world models built on top of it. Today, we’re also launching DOS 2.0 – our next-generation inference and training platform, delivering over 1,600 tokens per second for agentic inference and over 100 FPS for world models across major hardware platforms. Alongside DOS, we will launch new versions of our world models in the coming weeks: Lucy, for immersive realtime experiences, and Oasis, for physical AI. Grateful to our partners, customers, and backers across media & entertainment, Physical AI, chips, hyperscalers, and the broader AI ecosystem. Radical Ventures, Adobe, Alpha Partners, Amazon and Amazon Web Services (AWS), Atreides Management, LP, Benchmark, eBay Ventures, NVIDIA, Sequoia Capital, Toyota Ventures, Valley Capital Partners, Valor Equity Partners, Zeev Ventures; and Andrej Karpathy, Michael Eisner, Yamauchi-No.10 Family Office, Moritz Baier-Lentz and more. We made a film for this moment with Decart CEO Dean Leitersdorf and Moritz Baier-Lentz - a look at the company, the technology, and what comes next.
-
Excited to share that we’ve raised $300M in our Series B round, led by Radical Ventures and Jordan Jacobs, bringing our total funding to more than $450M, with leading technology companies joining as both customers and investors. We continue accelerating the path to AGI through our two core pillars: ultra-optimized infrastructure for AI workloads, and realtime world models built on top of it. Today, we’re also launching DOS 2.0 – our next-generation inference and training platform, delivering over 1,600 tokens per second for agentic inference and over 100 FPS for world models across major hardware platforms. Alongside DOS, we will launch new versions of our world models in the coming weeks: Lucy, for immersive realtime experiences, and Oasis, for physical AI. Grateful to our partners, customers, and backers across media & entertainment, Physical AI, chips, hyperscalers, and the broader AI ecosystem. Radical Ventures, Adobe, Alpha Partners, Amazon and Amazon Web Services (AWS), Atreides Management, LP, Benchmark, eBay Ventures, NVIDIA, Sequoia Capital, Toyota Ventures, Valley Capital Partners, Valor Equity Partners, Zeev Ventures; and Andrej Karpathy, Michael Eisner, Yamauchi-No.10 Family Office, Moritz Baier-Lentz and more. We made a film for this moment with Decart CEO Dean Leitersdorf and Moritz Baier-Lentz - a look at the company, the technology, and what comes next.
-
Love seeing loook.ai use Decart to bring realtime AI experiences into the physical world through their Smart Mirror Platform.
World’s first real-time AI clothing try-on for H&M Group! We powered the H&M Group x Stella McCartney collection launch event in Lima, Peru, enabling guests to virtually try on 40+ looks from the collection in real time. +260 minutes of active usage during the event, hundreds of virtual try-ons and social media shares! Developed in collaboration with Decart and Proyecto Tregua. Stay tuned for more updates! #ai #smartmirror #retail #tech
-
We continue to improve Lucy, focused on quality, stability, and control, while maintaining realtime performance. This update includes more accurate lip sync, sharper visuals, and more natural lighting, along with greater stability and better handling of complex characters and motion. When high-quality video is generated in realtime, entire categories shift - from shopping and advertising to streaming, gaming, entertainment, and sports. Build the future → https://lnkd.in/duPUhMP4
-
What if video editing had a “volume knob”? We are excited to share new research from Decart and Tel Aviv University, accepted to #SIGGRAPH2026. We found directions within video models that enable smooth control over edit strength, from subtle changes to full transformations! Why is controlling edit strength hard? In classifier-free guidance, lowering the guidance scale does not smoothly reduce the edit. It often moves the model toward an arbitrary “null” edit instead of true reconstruction, which leads to unpredictable behavior. We fix this by redefining the guidance origin. Instead of interpolating toward a null-conditioned prediction, we adapt the origin toward an identity direction. This turns edit intensity into a much smoother and more semantic control signal for video editing. Want to dive deeper? Check out our blog post, paper and some more examples in the comments. See you at SIGGRAPH 2026!
-
LIVE from HumanX Conference 2026 in San Francisco - Decart CEO Dean Leitersdorf on stage, sharing how realtime AI is redefining the boundary between watching and experiencing.