Releases

One year after launch, Tinker had expanded from 14 text-only models to more than 20 models spanning 4B to 1T parameters, added multimodal and audio support, enterprise controls, full fine-tuning beta, and production-inference compatibility.

An independent careers index counted 53 live roles on Thinking Machines Lab's verified Ashby hiring board. The number is a point-in-time external snapshot.

Thinking Machines and collaborators reported that ReViSQL-K2.6 reached 92.97% accuracy on Arcwise-Plat-SQL with 16-sample self-consistency, narrowly exceeding the 92.96% human proxy at a cost of $0.56 per task.

Thinking Machines opened a grant program offering up to $50,000 in Tinker credits for safety research on open-weight models, including alignment failures, risk measurement and safer model opening.

Thinking Machines released Inkling-Small, a 276B-total-parameter mixture-of-experts model with 12B active parameters and performance comparable to the larger Inkling model.

Thinking Machines published Inkling's full weights and made the model available for customization through Tinker.

Thinking Machines released Inkling, a 975B-total-parameter multimodal mixture-of-experts model with 41B active parameters and up to a one-million-token context window.

Thinking Machines published research on training language models to replicate expert judgment in financial tasks using Tinker-based post-training.

Thinking Machines was reported to have grown from about 30 employees to roughly 120 over its first year.

Thinking Machines and NVIDIA announced a multi-year partnership to deploy at least one gigawatt of Vera Rubin systems, alongside an NVIDIA investment.

Tinker became generally available with vision input, Kimi K2 Thinking support and an OpenAI-compatible sampling interface.

The company released the open-source Tinker Cookbook with modern post-training implementations and examples built on the Tinker API.

Thinking Machines launched Tinker, a managed API that exposes low-level primitives for fine-tuning and post-training open-weight models.

Thinking Machines Lab closed a $2 billion seed financing led by Andreessen Horowitz with participation from NVIDIA, Accel, AMD, Cisco, ServiceNow and Jane Street.

At launch, Thinking Machines Lab was reported to have approximately 30 employees, including a number of former OpenAI researchers and engineers.

Mira Murati publicly introduced Thinking Machines Lab and its mission to build AI systems that are more understandable, customizable and collaborative.