Inception's model writes its whole answer at once, not word by word.
The California startup Inception released Mercury 2.5 on Tuesday, a model that writes text the way image tools paint. It clocks 1,107 tokens a second, where a token is a chunk of a few letters. No independent tester has scored it yet, so the quality claims are still the company's own.
Why it mattersSpeed decides whether a voice assistant feels alive or leaves you sitting through an awkward pause. A fast cheap model gives builders an option outside the big labs for everyday jobs.
Read in the app — free, in 9 languages
Related stories
DeepSeek will route its top model's users to a cheaper one.2026-09-09A tiny open Chinese model matched one four times its size.
2026-09-09iFlytek trained a 293-billion-parameter model without Nvidia chips.
2026-09-07Outside testers scored OpenAI's new Astra below Anthropic's Fable 5.1.
2026-09-05A Chinese lab built Saudi Arabia's flagship Arabic AI model.
2026-09-04