Thinking Machines ships Inkling 2 Small, weights due next week
A 594B mixture-of-experts model with a vision encoder, available now on Tinker for inference and fine-tuning, with Apache 2.0 weights planned once safety testing finishes.
Thinking Machines Lab released Inkling 2 Small on 5 November, a mixture-of-experts model with 594 billion total parameters and 17 billion active, plus a 292-million-parameter vision encoder for image input. It is live now on the company’s Tinker platform for inference and fine-tuning.
The model was pretrained on 8.4 trillion tokens across four phases, then put through a joint reinforcement-learning run over nine days with more than 17 million rollouts. Thinking Machines reports a score of 31 on the Artificial Analysis Intelligence Index, 72.1% on DeepSWE v1.1 and 86.3% on the hard split of SWE-Bench Pro v2, enough to put it among the strongest open-weights models outside China. Full weights under an Apache 2.0 licence are planned for the following week, once external safety testing finishes.
Why it matters: this is Thinking Machines’ second model after Tinker itself, and the first with vision input. If the promised Apache 2.0 weights land as described, it is another serious self-hostable option alongside Qwen, GLM and DeepSeek, from a lab with no frontier release before this year.
The caveat: every benchmark here is the vendor’s own, and as of this release the weights are not out, only Tinker access. There is nothing to independently verify yet.