Thinking Machines ships a 594B model running 17B active on Tinker

single source· 1 articles · confidence: low · first seen 2026-11-06 04:25 UTC

What this means for you

Nothing to act on yet. The weights are only planned, so Tinker is currently the single way to reach this model, and no price was published. Treat the 31 as provisional until you know what was run and when.

Thinking Machines Lab has released Inkling 2 Small, a mixture-of-experts model (only part of the network runs on each request, so serving cost tracks the smaller number) with 594bn total parameters and 17bn active. It runs on Tinker, the company's training and inference service; Apache 2.0 weights are described as planned, not published. It scores 31 on the Artificial Analysis Intelligence Index, reported without an evaluation date or harness, so the figure cannot be lined up against older scores without checking. No price was given.

Models in this story

Key facts

  • ·Inkling 2 Small has 594bn total parameters with 17bn active per request. source
  • ·It scores 31 on the Artificial Analysis Intelligence Index, with no evaluation date or harness given. source
  • ·The model runs on Tinker, Thinking Machines Lab's training and inference service. source
  • ·Apache 2.0 weights are planned rather than released. source
  • ·The release was reported on 6 November 2026. source

What the sources say

  • MarkTechPost — Short release note carrying the parameter split, one index score and the licensing plan.

Sources

The original reporting. Follow these — they did the work.

← the wire