Thinking Machines Releases Inkling-Small, a 276B Open-Weight Model Within One Point of Its Flagship

Inkling-Small is a 276-billion-parameter MoE model with only 12 billion active parameters, multimodal inputs across text, image, and audio, and a 1M-token context window, all under Apache 2.0. It scores 80.2 percent on SWE-bench Verified, narrowly topping the 975-billion-parameter flagship, and runs at roughly a quarter of the inference cost.