Thinking Machines Lab introduced Inkling 2 Small on 5 November 2026, its second-generation model. The sparse mixture-of-experts system has 594 billion total parameters with 17 billion active per token, and it targets long-horizon autonomous work alongside coding, vision and knowledge tasks. At launch it is available only through the lab's Tinker platform — inference in the Tinker Playground, plus fine-tuning and sampling on Tinker — with the weights still closed.

The headline number is the Artificial Analysis Intelligence Index score of 31, up 18 points from the first-generation Inkling Small. Artificial Analysis says that once the weights are released, Inkling 2 Small would be the highest-scoring open-weights model developed outside China. The model is a 48-layer decoder-only transformer routing each token to eight of 384 experts, with a 1M-token context, and Tinker pricing of $0.30 per million input tokens and $1.00 per million output tokens.

One honesty note matters: the "open weights" part of the launch is a plan, not a delivery. Thinking Machines says it intends to publish the full weights under Apache 2.0 next week, but only after external safety testing and red-teaming is complete. The lab's staged-release framework makes that progression evidence-dependent rather than automatic, so the download date is conditional, not guaranteed. Until then, nobody can self-host or download the model — Tinker is the only access route.

The launch still matters. A 594B model with a 31 Intelligence Index score from Mira Murati's lab, arriving the same week as Tencent's Hy4 and OpenAI's GPT-6 Cyber, shows how fast the frontier is moving — and how much of it is now gated behind hosted platforms first. If the weights land next week as planned, Inkling 2 Small becomes a serious open-weights option; if safety testing drags on, the launch reads as another closed release with a promise attached.

Sources