Thinking Machines launched open-weight Inkling-Small

Thinking Machines Lab released Inkling-Small, an open-weight Mixture-of-Experts model that aims to match Inkling’s performance at about one-quarter the size.

The full weights are now available on Hugging Face, with fine-tuning and chat access through Tinker and Tinker Playground.

Inkling-Small offers a one-million-token context window, variable thinking effort, multimodal support, and lower training and inference compute.

The lab reported strong results on reasoning, coding, and agentic benchmarks, while noting some tradeoffs in factuality and knowledge coverage versus Inkling.