Thinking Machines Lab launches Inkling 2 Small on Tinker

Thinking Machines Lab announced on X on 5 November 2026 that Inkling 2 Small is available for fine-tuning and inference on Tinker, with a full weight release planned soon. The verified company account wrote:

“Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it’s a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon.”

Co-founder Mira Murati quoted that post, saying the model is “available today on Tinker, with weights coming soon,” and calling it “a capable small model.” Co-founder John Schulman separately posted that the accompanying blog includes detail on the training process and alignment sections.

What the company blog says

The Inkling 2 Small announcement dated 5 November 2026 describes a Mixture-of-Experts model with 594 billion total parameters and 17 billion active per token, with image inputs and variable effort. Thinking Machines says the model scores 31 on the Artificial Analysis Intelligence Index and will initially ship through Tinker with closed weights, with full weights intended under an Apache 2.0 license “next week” after external safety testing and red-teaming.

The post attributes capability gains to more efficient pre-training and a large unified reinforcement-learning run. It also states that early checkpoints were tested on OpenCode under the name Fledge-Alpha. Benchmark tables, safety scores and training FLOPs figures in the post are the company’s own reported numbers; this article does not independently re-run those evaluations.

Caveats

Availability, licensing timing, parameter counts and benchmark claims come from Thinking Machines’ X posts and company blog dated 5–6 November 2026. Treat the planned Apache 2.0 weight release as contingent on the safety testing the company says is still underway.

This report is not legal, financial, or health advice. See the disclaimer.