Thinking Machines releases Inkling 2 Small, a 594-billion-parameter model it ranks first among Western open-weight models

Thinking Machines, the startup led by former OpenAI CTO Mira Murati, has released Inkling 2 Small, its second generation of models, The Decoder reports. It is a mixture-of-experts model with 594 billion parameters in total, of which 17 billion are active at any one time. Inkling 2 Small scores 31 on the Artificial Analysis Intelligence Index. That is 18 points higher than its predecessor and roughly level with Google's Gemini 3.8 Flash, with especially large gains on agentic tasks. The gains come at a price. The model produces more output tokens and costs six times as much per task as the previous version. Inkling models are designed as base models for fine-tuning, which could make them useful for specialized applications. At first, Inkling 2 Small is available only as a proprietary offering through Thinking Machines' Tinker platform. The company says the weights will be released later under an Apache license. Until that happens, the model's claim to be the strongest Western open-weight model rests on a promised release. For developers who want to fine-tune their own systems, it would be a capable new base model, though one with notably higher running costs than its predecessor.