Thinking Machines releases Inkling, its first open model, betting against one-size-fits-all AI

Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling

Thinking Machines Lab, founded by ex-OpenAI CTO Mira Murati, released Inkling, a 975B-parameter mixture-of-experts model that activates only ~30B per inference. Weights are fully open for download and modification. The company explicitly rejects the one-giant-model approach and bets on multi-model collaboration. A technical report and weights are available, but training cost, pricing, and API timeline are not disclosed. Open weights are a good signal, but real-world performance still needs third-party benchmarks.

Why it matters: First public release from Mira Murati's lab: 975B MoE with open weights and a clear anti-monolith stance. HKR all hit. Held below 85 because training cost, API pricing, and timeline are all undisclosed — real-world usability is still an open question.