Updates from Inception Labs
Building the next generation of diffusion-based large language models for real-time production applications.
- Introducing Mercury, our General Chat Diffusion Large Language Model (Product launch · October 19, 2026)
Inception released Mercury 3, a flagship reasoning model featuring a one-million-token context window and parallel token generation for improved throughput and cost efficiency. - Mercury 2: the first reasoning model fast enough to pick up the phone (Product launch · October 15, 2026)
The company introduced Mercury 2, a reasoning diffusion language model capable of decoding over 1,000 tokens per second to support real-time voice applications.