Thinking Machines unveils Inkling 2 Small open-weight model

Thinking Machines, the startup founded by former OpenAI CTO Mira Murati, has introduced Inkling 2 Small, a 594-billion-parameter mixture-of-experts model. About 17 billion parameters are active at a time. The model accepts text and images and allows users to adjust its thinking depth. Thinking Machines describes it as the strongest Western open-weight model, although it is initially available as a proprietary service through the Tinker platform.

The company says Inkling 2 Small scored 31 on the Artificial Analysis Intelligence Index, compared with 13 for Inkling Small and roughly matching Google’s Gemini 3.8 Flash. Its largest gains were in agentic tasks: AutomationBench-AA rose from 5% to 46%, while FrontierCode 2.0 increased from 5% to 25%. On AA-Briefcase, it reached 1,527 Elo. The reported hallucination rate fell to 27% from 61%, alongside a lower response rate of 53% instead of 77%.

The model uses about 150,000 output tokens per task on average and costs $1.18 per task, nearly six times more than its predecessor. Thinking Machines plans to publish the weights under an Apache 2.0 license after external safety testing. The company says the model does not support audio and can be inefficient with tools and output.