Thinking Machines Lab Reveals Inkling 2 Small, a 594-Billion-Parameter Open Model
Thinking Machines Lab, the AI research company founded by former OpenAI CTO Mira Murati, announced Inkling 2 Small on November 5, 2026. The mixture-of-experts model has 594 billion total parameters and 17 billion active, and accepts text and image input.
1 source 1h ago
Reporting from 1 source: GIGAZINE.
Why it matters
The scores matter less than the plan to publish weights: an open model at Gemini 3.8 Flash's level would be, in the article's phrase, "the highest-performing open model from a company outside China."
Key Facts
- Mira Murati's Thinking Machines Lab announced the Inkling 2 Small model on November 5, 2026.
- Inkling 2 Small has 594 billion total parameters and 17 billion active parameters, and accepts text and image input.
- Thinking Machines Lab plans to release Inkling 2 Small as an open model in the second week of November 2026, after safety evaluations.
- Inkling 2 Small scored 31 on the Artificial Analysis Intelligence Index v5.0.1, tying Gemini 3.8 Flash and up from 13 for the previous generation.
- DesignArena testing placed Inkling 2 Small above GPT-6.1 Sol.
Safety evaluations are still running, and the company plans to release it as an open model in the second week of November 2026. It scored 31 on the Artificial Analysis Intelligence Index v5.0.1, tying Gemini 3.8 Flash and up from 13 for the previous generation.
On DesignArena, which measures design performance, Inkling 2 Small scored above GPT-6.1 Sol. The published comparisons also set it against GLM-5.3 Flash, DeepSeek-V4.1-Flash-1017, Muse Spark Open and Kimi K3 on coding and autonomous research tasks, where it reaches scores equivalent to cutting-edge Chinese-made models.
The other pitch is headroom. Thinking Machines Lab's graphs show more performance gained per training step and per unit of training cost than DeepSeek-V4.1-Flash-1017, meaning further training buys more for the money. A model card with the full specifications is published alongside the announcement.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google