Thinking Machines releases Inkling 2 Small: AA index 31 but 86.3% on SWE-Bench Pro, open weights under Apache 2.0 next week
First seen on X 2 days ago@thinkymachines ♥ 4,8924 views
"Scoring 31 on the AA Intelligence Index, a big jump over its predecessor." Thinking Machines' Nov. 5 announcement passed about 4,900 likes and 1.3 million views. The score alone is not flashy, so we checked the blog to see why it spread.
According to the Thinking Machines blog, Inkling 2 Small is a mixture-of-experts model with 594 billion total parameters and 17 billion active, and it accepts image input. Its table lists an Artificial Analysis Intelligence Index v5.0 score of 30.6. On coding it reports 72.1% on DeepSWE v1.1, 86.3% on SWE-Bench Pro v2 (Hard) and 81.6% on Vibe Code Bench v1.1. In the index comparison, Claude Haiku 5.5 is highest at 36.5, with DeepSeek-V4.1-Flash at 32.6, GLM 5.3 at 31.7 and Gemini 3.8 Flash at 30.8, all ahead of Inkling.
The model is available on Thinking Machines' Tinker with closed weights for now. The company says it will release full weights under Apache 2.0 next week after external safety testing, and other platforms will follow "in the coming weeks." It also listed limits: chat replies are less succinct and polished, it sometimes picks poor tools in codebases with many skills, and it overreacts with disclaimers on mental-health questions.
Reaction is mixed. One X user called 31 the weakest major open release of the quarter and told readers to look at price. Another claimed it ranks third on DesignArena without a source. Some read the interest as being about the promise to release weights rather than the index rank.
“Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor.”
