Thinking Machines Lab launches Inkling 2 Small, a 594B-parameter sparse MoE model, on Tinker

Thinking Machines Lab introduced Inkling 2 Small on November 5, 2026, making its second-generation model available through Tinker with closed weights. The model is a sparse Mixture-of-Experts system with 594 billion total parameters and 17 billion active parameters.

Key specifications

• Architecture: sparse Mixture-of-Experts
• Total parameters: 594 billion
• Active parameters: 17 billion
• Weights: closed at launch (the headline on Unite.AI's report describes open weights as pending)
• Context window: not yet disclosed
• Pricing: not yet disclosed
• Benchmark scores: not yet disclosed

Focus areas

Thinking Machines Lab targets long-horizon autonomous work with Inkling 2 Small, alongside coding, vision and knowledge tasks.

Access

At launch, Tinker Playground provides inference-only access to the model. Tinker supports fine-tuning and sampling under the platform's terms.

The details above come from a short summary published by Unite.AI.

What this means

Inkling 2 Small is available through Tinker, where users can fine-tune and sample from it, while Tinker Playground is limited to inference. Weights are closed at launch, so Tinker is the access path for now, and Unite.AI's headline indicates an open-weights release is still pending.