Thinking Machines launches Inkling 2 Small
Thinking Machines' Inkling 2 Small scores level with Gemini 3.8 Flash on Artificial Analysis's index, and its promised weights release would make it the top-scoring open-weights model built outside China.
Thinking Machines has released Inkling 2 Small. Artificial Analysis reports that it scores 31 on its Intelligence Index v5.0.1. That is 18 points above Inkling Small, level with Google DeepMind's Gemini 3.8 Flash (high) and ahead of Xiaomi's MiMo-V2.6-Pro.
For now the model is proprietary. It is available only through Thinking Machines' Tinker platform and Tinker Playground. Weights are planned under Apache 2.0, but no date has been confirmed. It is a reasoning model with 594B total and 17B active parameters. It takes text and image input and produces text output. Pricing is $0.30/$1.00 per 1M input/output tokens.
The biggest gains are on agentic tasks. AutomationBench-AA rose from 5% to 46% and FrontierCode 2.0 from 5% to 25%. On FrontierCode 2.0 the model beats Claude Sonnet 5.5 (medium), which scores 11%. It trails Sonnet on Terminal-Bench 4.0, 13% to 30%. On AA-Briefcase, a test of long-horizon business workflows, it reached 1527 Elo, compared with 1439 for Sonnet 5.5 and 1203 for Gemini 3.8 Flash.
The hallucination rate on AA-Omniscience fell from 61% to 27%. Artificial Analysis attributes this mainly to the model attempting fewer questions (53% vs. 77%). Accuracy rose only from 32% to 34%.
The gains come with much heavier output. The model averages ~150K output tokens per index task, roughly 4x Inkling Small. Cost per task rose almost 6x, to $1.18. That is still well below Gemini 3.8 Flash (high) at $7.77 and Claude Sonnet 5.5 (medium) at $10.02.