Inkling

Inkling is Thinking Machines Lab's 975B-parameter open-weights Mixture-of-Experts model, released under Apache 2.0 on 2026-07-21. It accepts text, image, and audio input and generates text, with native reasoning and variable thinking effort (the published figures use effort=0.99). The vendor model card reports 97.1% on AIME 2026, 77.6% on SWE-Bench Verified, 73.5% on MMMU-Pro, 77.2% on audio MMAU, and 91.4% on the VoiceBench average, positioning it as the flagship sibling of the smaller Inkling-Small. It is served by DeepInfra with a 512K-token context window at $0.95 / $4.05 per 1M input/output tokens.