Models New
Inception releases Mercury 3, a diffusion language model it says runs at more than 1,000 tokens per second
Published:
Inception announced Mercury 3 on 19 October and called it the most capable diffusion-based language model released to date. Diffusion models generate many tokens in parallel through repeated denoising, rather than one token at a time.
The company says Mercury 3 is more than 40 times larger than Mercury 1, runs at 1,068 tokens per second on Nvidia GPUs and has a one-million-token context window, up from 260,000 for Mercury 2.5. It claims quality on par with or better than speed-optimised models such as GPT-6 Luna and Claude Haiku 5.5, while running four to eight times faster.
List prices are $0.20 per million input tokens and $0.70 per million output tokens, with a 50% launch discount to $0.10 and $0.35. The model is available through Inception's own API, OpenRouter, Baseten and Vercel, and the API is compatible with OpenAI's.
Caveat. All benchmark and speed figures come from Inception, and this edition found no independent test. Inception has not published the parameter count or said whether it will release the weights. Both questions were asked in the launch thread.
Sources