Inception (Inception Labs)
Diffusion language models (the Mercury family) generate text by iterative refinement, with far lower latency and cost than token-by-token generation. Mercury 2 reportedly runs about 1,000 tokens per second.
- Founded
- 2024
- Headquarters
- Palo Alto, CA
- Founders
- Stefano Ermon, Aditya Grover, Volodymyr Kuleshov
- Pushes against
- Autoregressive next-token transformers.
- Now
- Reportedly drew acquisition interest from Microsoft and SpaceX around 2026-05-13 and sought over $1B (Reuters via BigGo); Mercury 2 launched 2026-02-24 per a Business Wire headline. Unconfirmed whether any deal closed.
Key people
- Stefano Ermon (Stanford professor, diffusion-model pioneer)
- Aditya Grover (UCLA professor, co-lead)
- Volodymyr Kuleshov (Cornell professor, co-lead)
Funding
- 6 November 2025
- Amount
- $50M seed
- Lead
- Menlo Ventures (Mayfield, Innovation Endeavors, Microsoft M12, Snowflake, Databricks, NVentures; angels Andrew Ng and Andrej Karpathy)
Every funding round in the atlas
Launches
- Mercury (diffusion LLMs) Inception Labs · 17 June 2025
Mercury offers commercial-scale diffusion LLMs for code, and Mercury Coder Mini runs at 1,109 tokens/s on H100, 10x faster than speed-optimised frontier models at similar quality.
Sources
- techcrunch.com/2025/11/06/inception-raises-50-million-to-build-diffusion-models-for-code-a
- finance.biggo.com/news/UZn6OZ4BoQmpnl36fQqH
- techcrunch.com/2025/02/26/inception-emerges-from-stealth-with-a-new-type-of-ai-model/
- mayfield.com/introducing-inception/
- finance.yahoo.com/news/exclusive-microsoft-eyeing-startup-deals-201449688.html
- the-decoder.com/inception-launches-mercury-2-the-first-diffusion-based-language-reasoning-
This profile was checked and corrected against its sources on 6 October 2026. How we check