Thinking Machines Lab, founded by Mira Murati after leaving OpenAI in September 2024, released Inkling on July 15. This open-source model features 975 billion total parameters with 41 billion active during inference, making it the first major model trained from scratch by a Western lab. Its performance on MCP Atlas (74.1%) and SWE-Bench Verified (77.6%) surpasses Nvidia’s Nemotron 3 Ultra, though it remains outperformed by Chinese models on raw coding benchmarks. The model is available on OpenRouter at $1 per million input tokens and $4.05 per million output tokens, with notable censorship that limits certain use cases.
Source: Read the original article
