The short answer
● Unveiled — 6 November, Thinking Machines
● Size — 594B total, 17B active, mixture of experts
● Intelligence Index — 31, up from 13 for its predecessor
● Cost per task — $1.18, against $7.77 for Gemini 3.8 Flash
● The catch — the weights are not out yet
Thinking Machines unveiled Inkling 2 Small on 6 November: a 594-billion-parameter mixture of experts with 17 billion active, accepting text and image input. Its Artificial Analysis Intelligence Index score is 31, up from 13 for Inkling Small, which puts it roughly level with Gemini 3.8 Flash.
The economics are the headline. Artificial Analysis puts it at about 150,000 output tokens per task, roughly four times its predecessor, at $1.18 per task against $7.77 for Gemini 3.8 Flash and $10.02 for Claude Sonnet 5.5. On coding it scores 25.4% on FrontierCode v2.0, 72.1% on DeepSWE and 86.3% on SWEBench Pro v2 Hard, where Kimi K3 narrowly leads.
It is being reported as an open-weight release and it is not one yet. The model runs today on Thinking Machines’ Tinker platform as a proprietary offering. Apache 2.0 weights are promised once external safety testing finishes, with no date given. One more number worth reading carefully: the hallucination rate fell from 61% to 27%, but the response rate fell from 77% to 53%, so some of that improvement is the model declining to answer.
Sources: The Decoder · MarkTechPost.
Published 8 November 2026. Every figure above was checked against the reporting linked here on the same day.