16 October 2026
Inception releases Mercury 3, a faster diffusion-style model, claiming benchmark wins
First reported
Deep Learning Weekly ran this on , 4 days before the next source picked it up.
- Inception, a startup co-founded by Stefano Ermon, released Mercury 3, which it says is faster than rival chatbots.
- The model drafts whole answers at once and refines them in parallel, rather than writing one word at a time.
- In the company's own tests it generated about 1,068 words' worth of text per second, versus about 240 for Anthropic's Haiku 5.5 and 130 for OpenAI's GPT-6 Luna.
Our read
Only Deep Learning Weekly among the 30 newsletters covered this release, with SiliconANGLE also reporting it. Our archive has an earlier story on the same thread, Inception's Mercury 3 launch, dated 2026-10-20.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
Deep Learning WeeklyEditorial team
Pricing tiers, Terminal-Bench 4.0 score, and comparison with GPT-6 Luna
Reported by SiliconANGLE