Inception unveils Mercury 3 diffusion LLM with 1,000 tps
According to The Rundown AI citing Inception, Mercury 3 is a 40x larger diffusion language model than Mercury 1 that delivers 1,000+ tokens per second, matches or beats GPT-6 Luna and Claude Haiku 5.5, and targets low-latency use cases.
SourceAnalysis
On October 19, 2026, Inception launched Mercury 3, which the company describes as the strongest diffusion language model released to date according to The Rundown AI.
The model is more than 40 times larger than Mercury 1 while maintaining output speeds above 1,000 tokens per second and a 1 million token context window. Inception states that Mercury 3 matches or exceeds the performance of OpenAI GPT-6 Luna and Anthropic Claude Haiku 5.5 on standard benchmarks in the fast and affordable tier. Diffusion models generate text by drafting multiple words simultaneously and then refining them, unlike the sequential word-by-word approach used by ChatGPT and Claude, which suits applications such as voice agents, live tutoring, and customer support where low latency matters. Mercury 3 is available immediately through the Inception API, Baseten, and OpenRouter.
The Rundown AI
@TheRundownAIUpdating the world’s largest AI newsletter keeping 2,000,000+ daily readers ahead of the curve. Get the latest AI news and how to apply it in 5 minutes.