Inception’s Mercury 3 Brings a Diffusion Model to the Fast Tier
More than 40x the size of Mercury 1, over 1,000 tokens a second and a 1M-token context window.
A diffusion language model now claims the fast, cheap tier.
Inception has launched Mercury 3. The company calls it “the most capable diffusion LLM released to date” and says it outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks while running at over 1,000 tokens per second.
| Measure | Mercury 3 |
|---|---|
| Size | More than 40x Mercury 1 |
| Output speed | Over 1,000 tokens per second |
| Standard benchmarks | 3x more intelligent than Mercury 1 |
| Context window | 1M tokens |
| Compared against | GPT-6 Luna, Claude Haiku 5.5 |
| Available on | Inception API, Baseten, OpenRouter |
Inception says Mercury 3 is “just as fast” as its predecessor despite the jump in size. The Rundown AI puts it more plainly: 40x bigger than Mercury 1, “but still puts out 1,000+ tokens a second.”
Why diffusion matters for speed.
The Rundown AI explains the difference. ChatGPT and Claude write one word at a time. Diffusion models draft many words at once and then refine them.
That makes them a strong fit, The Rundown says, for voice agents, live tutoring and customer support, “where any lag shows.” The Rundown also notes the comparison Inception draws is with the fast, cheap tier of OpenAI and Anthropic models.
Sources
- @TheRundownAI on X · Oct 19, 11:54 AM PT