Thinking Machines Lab is a Frontier Foundation Labs company founded in 2025 and based in San Francisco, United States. It has raised $2B in total funding, most recently a Seed in 2025 at a $12B valuation.
United States| Date | Stage | Amount | Valuation | Lead investors |
|---|---|---|---|---|
| Jul 15, 2025 | Seed | $2B | $12B | Andreessen Horowitz |
Thinking Machines Lab converted Mira Murati's OpenAI pedigree into a $2B seed round — capital raised on team alone — and its first shipped product, Tinker, stakes out the post-training layer: making fine-tuning of open-weight models simple. It is a coherent wedge, but the moat question is unforgiving: fine-tuning services are offered by Together, Fireworks, and increasingly the model labs themselves, and nothing about an API launched in late 2025 has had time to accumulate switching costs, pricing history, or disclosed revenue. The scorecard is therefore mostly N/A with a C-shaped outline — extraordinary inputs, with the defensible business still to be demonstrated.
Mira Murati is the founder and CEO of Thinking Machines Lab, which she launched in February 2025. She was previously the chief technology officer of OpenAI, where she led the development of ChatGPT, DALL-E, and the GPT model series, and briefly served as interim CEO during the November 2023 leadership crisis. She founded Thinking Machines Lab to build frontier AI systems and tools for adapting them to specific needs.
John Schulman is a co-founder and chief scientist of Thinking Machines Lab. He was a co-founder of OpenAI, where he co-created Proximal Policy Optimization (PPO) and led reinforcement-learning and RLHF work behind ChatGPT. After a brief stint at Anthropic, he joined Thinking Machines Lab and remained as its chief scientist amid later departures from the founding team.
Eiso Kant releases Laguna S 2.1, claiming it outperforms DeepSeek V4 Pro with a 10x smaller model at lower cost than V4 Flash.
Thinking Machines Lab positions itself as a rival to Anthropic in the serious-AI-research space.
Thinking Machines Lab releases Inkling, a 975B-parameter multimodal MoE open model with 1M-token context, alongside a 276B Inkling-Small variant.
Thinking Machines Lab releases Inkling, a new open model hosted on Hugging Face.
An API for fine-tuning open-weight language models, launched October 1, 2025. It lets developers customize and post-train models for specific tasks or industries without managing the cost and complexity of distributed training infrastructure.
