Thinking Machines Lab releases Inkling 2 Small: a 594B MoE, closed weights on Tinker first, Apache 2.0 weights promised next week
Thinking Machines Lab released Inkling 2 Small on 5 November, a Mixture-of-Experts model with 594B total and 17B active parameters, image input and variable reasoning effort. It is available first on the company’s Tinker platform with closed weights; the company says Apache 2.0 weights will follow next week once external safety testing is complete.
According to the company’s blog, Inkling 2 Small scores 31 on the Artificial Analysis Intelligence Index, well above the company’s earlier models, with gains in agentic coding, vision, knowledge work and reasoning. The company credits improvements across training, including more efficient pre-training and a large unified reinforcement-learning run; RL covered hundreds of thousands of long-horizon coding tasks across several agent harnesses and context-compaction strategies, and some tasks used an agentic grader that judged how the model approached the problem, not only whether tests passed.
The model can be tried in Tinker Playground today, and Tinker offers fine-tuning and sampling; downloadable fine-tuned checkpoints arrive only once the weights are released. The company says the full weights will be released under Apache 2.0 next week after external safety testing and red-teaming, with more platforms to follow in the coming weeks. This follows the “safe path to open weights” approach it outlined earlier, so the release depends on the test results.
The company also lists limitations: no audio support; it is not the most succinct communicator and does not always pick the best tools, especially in codebases with many skills; and it is less optimized for chat, with poorer content organization, overuse of slashes and overreaction to mental-health and other safety-related queries. The capabilities and score are the vendor’s own claims.
Sources
- Introducing Inkling 2 Small — Thinking Machines Lab