Thinking Machines Lab set to close $5 billion Accel round at $40 billion valuation

Thinking Machines Lab is closing in on a $5 billion funding round at a $40 billion pre-money valuation, Business Insider reported on November 6. Accel is leading the round, Business Insider's sources familiar with the matter said. A representative for Thinking Machines Lab declined to comment, and Accel did not reply to a request for comment.

The new round would more than triple the startup's valuation from its last public round. Thinking Machines Lab raised a $2 billion seed round in July 2025 at a valuation above $10 billion from investors including Andreessen Horowitz, Nvidia, GV, Lightspeed Venture Partners, Jane Street and Factorial Funds. Last year, the company sought funding at a $50 billion valuation, but those talks were abandoned.

Thinking Machines Lab is led by Mira Murati, who was previously OpenAI's chief technology officer. Other key figures helped train the early versions of ChatGPT, and the company emerged from stealth last year as a high-profile AI lab.

The company's first tool, Tinker, lets users fine-tune large language models. On Thursday, Thinking Machines Lab released a new version of Tinker and said the model improved on AI coding. SiliconANGLE reported that the company also debuted Inkling 2 Small, a large language model that can automate AI training tasks.

Inkling 2 Small was trained on a cluster of GB300 NVL72 racks, each containing 72 Blackwell Ultra graphics cards and half as many central processing units. The company finished the initial pre-training phase in 68 hours and then ran a nine-day reinforcement learning workflow. The model has 594 billion parameters, with 292 million in a vision encoder and about 17 billion active when a user enters a prompt. It can adjust how much computing power it spends on tasks.

Thinking Machines tested Inkling 2 Small by having it fine-tune Qwen3.5-4B. The company says the model doubled the open-source LLM's forecasting score in 4.5 hours. Inkling 2 Small also surpassed Spark Open, Kimi K3 and several other open-source models on three knowledge work benchmarks and earned higher scores on financial task datasets.

Inkling 2 Small is available through the Tinker cloud service, and the company plans to release the model's weights under an open-source license next week, SiliconANGLE reported.

Thinking Machines Lab is raising the round amid intensifying competition for AI talent. Founding members Barret Zoph, Luke Metz and Andrew Tulloch have left for rival companies. In May, Business Insider reported that about a third of the founding team had left, driven partly by large pay packages from Meta and OpenAI. The company has also hired competitive coder Neal Wu.