Murati's model would top the open-weights chart — but you can't download it yet

Thinking Machines Lab released Inkling 2 Small on November 5, a 594-billion-parameter mixture-of-experts model with 17 billion active per token. It scores 31 on the Artificial Analysis Intelligence Index, up from 12 for the first Inkling, matching Google DeepMind's Gemini 3.8 Flash and passing Xiaomi's MiMo-V2.6-Pro.

The catch is the weights: at launch the model is closed, available only through the company's Tinker platform. Full weights are promised under an Apache 2.0 license next week, after external safety testing and red-teaming — and the company stresses that once weights are published, there is no taking them back.

The model is built for long autonomous work, scoring 86.3% on SWE-Bench Pro Hard. Its hallucination rate fell to 27% from 61% on one test, mostly because it now answers fewer questions — accuracy barely moved, from 32% to 34%.