Inception launches Mercury 3, its fastest diffusion LLM yet, with 1M-token context and 1,000+ tokens per second

Full report

Inception announced Mercury 3 in a press release on October 19. The model generates 1,068 tokens per second on NVIDIA GPUs, keeping it above the 1,000-tokens-per-second bar the Mercury line has been known for.

The context window grows to one million tokens, nearly four times Mercury 2.5's 260,000. Inception says Mercury 3 is more than 40 times the size of Mercury 1 in total parameters, with intelligence across a range of standard benchmarks roughly tripling from Mercury 1 (mini) to Mercury 3. These are the company's own comparisons; independent evaluations have not yet been published.

Standard pricing is $0.20 per million input tokens and $0.70 per million output tokens, with launch pricing of $0.10 and $0.35. Cached input tokens are listed at $0.02 per million ($0.01 under the promotion).

Mercury 3 is available through the Inception API, OpenRouter, Baseten, and Vercel, and supports the OpenAI API format. Early customers are already using it for browser-use agents, interactive education, real-time search agents, and customer support, the company says.

Co-founder Stefano Ermon said the question heard most since Mercury 1 has been whether diffusion can scale. Mercury 3's answer: the limits to scaling diffusion models are compute and data, not fundamental algorithmic barriers. He added that training of the next model has already begun.

This is Inception's fourth release since Mercury 1 in June 2025: Mercury 2 introduced reasoning to diffusion models, Mercury 2.5 added tunable reasoning and a larger context window, and Mercury 3 delivers the largest jump in intelligence of any Mercury release.