Apple

Thinking Machines’ new model takes the lead in western open-weight AI

ZDNET’s key takeaways

Thinking Machines released a competitive new open model, Inkling 2 Small.
Most competitive open models come out of China.
The release follows French lab Mistral’s Le Chonk open model, which excels at coding.

Thinking Machines Lab, a small AI startup from ex-OpenAI CTO Mira Murati, has released its latest model, Inkling 2 Small — and according to third-party tester Artificial Analysis, it’s set to give Chinese models a run for their lead. It’s technically proprietary until now, but Thinking Machines said it will release its weights next week.

Also: Open weights vs. closed: An AI civil war’s afoot

Chinese companies have led the pack on open models over the last few years. Models from DeepSeek and Moonshot regularly top leaderboards and rival the capabilities of much more expensive, closed models from OpenAI and Anthropic.

What Inkling 2 Small does

Thinking Machines’ latest model has 594 billion parameters and supports image inputs but outputs only text. The company noted its web and game design capabilities and that it evaluated the model’s approach to coding problems alongside the code it generated during testing.

“The result is a model that is both stronger at coding and easier to use inside real agent loops,” the company said.

Also: Nvidia takes another step forward for US open-weight AI

Inkling 2 Small scored a 31 on the Artificial Analysis Intelligence Index, a composite test “for tracking progress toward artificial general intelligence” that combines benchmark scores spanning coding, science, and general categories. A score of 31 is above average for models in its class and is tied with Gemini 3.8 Flash. For context, Gemini 4 Carbon and Claude Opus 5.5 top the index at 60 and 59, respectively.

The open-weight race

Thinking Machines said it will release the model’s weights next week, a practice that is becoming customary. Given the heightened security concerns around AI in general, AI companies spend a few weeks testing the to-be-open model with cybersecurity teams and safety reviewers before releasing the weights, rather than leading with the weights out of the gate.

French AI lab Mistral recently released its Le Chonk model via the same staggered process — it now keeps pace with OpenAI’s GPT-6.1 Terra in coding abilities, which is rare for western-made open models.

Also: Mistral’s Le Chonk is now a top open model for coding

“[Thinking Machines’] plan to release its weights would make it the highest-scoring open weights model developed outside China based on current rankings,” Artificial Analysis said.

But rarely is a model the full package.

“Inkling 2 Small is amongst the leading models in intelligence and reasonably priced when comparing to other models of similar price. It’s also notably slow per task and very verbose,” Artificial Analysis said. The organization also called Inkling 2 Small’s pricing “somewhat expensive” at $0.30 per one million input tokens, since the median cost for its class is $0.24.

Still, the model’s progress over its predecessor stands out to Artificial Analysis reviewers: Inkling 2 Small jumped 41% in automation capabilities and 20% in coding abilities. However, most models excel in certain areas or on specific benchmarks, making it hard to deem one or another the outright best coding model overall.

Also: I tried replacing Claude Code with a local AI stack that’s completely free

Inkling 2 Small also made some gains in accuracy, Artificial Analysis said.

“On AA-Omniscience v1.1, the hallucination rate falls from Inkling Small’s 61% to 27%, while accuracy rises from 32% to 34% and the attempt rate falls from 77% to 53%,” its review reported.

Source: Read original story