Thinking Machines Lab, founded by ex-OpenAI CTO, launches Inkling AI model

Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has released its first AI model called Inkling. It's a multimodal model using a mixture-of-experts design, with 975 billion total parameters and 41 billion active. It handles text, images, audio, and video, and can process up to 1 million tokens of context. The model is now open-source on Hugging Face and available via their API.

Compared to other open-source models, Inkling shines in math, scoring 97.1% on the AIME 2026 test, beating DeepSeek V4 Pro's 96.7%. But it falls behind in coding and reasoning. Chinese open-source models like GLM 5.2 and DeepSeek V4 Pro outperform Inkling in these areas. For example, GLM 5.2 got 62.1% on SWEBench Pro, while Inkling got 54.3%.

Against US rival Nemotron 3 Ultra, Inkling wins in reasoning, coding, and agent workflows. However, compared to closed-source giants like Claude Fable 5, GPT 5.6 Sol, and Gemini 3.1 Pro, Inkling lags in extreme reasoning and software engineering. Still, it's competitive in multimodal tasks, especially audio processing, where it nearly matches top models.