Thinking Machines Unveils Inkling 2 Small, Challenging AI Rivals

Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has introduced its latest AI model named Inkling 2 Small. This model aims to disrupt the current market dynamics dominated by Asian competitors, particularly in the open-weight AI sector.

Performance Specifications

The Inkling 2 Small model features a staggering 594 billion parameters, with specifications that include the ability to process image inputs while limiting its outputs to text. The developers have particularly highlighted its strengths in web and game design, as well as its efficient handling of coding tasks.

Competitive Landscape

In terms of performance, Inkling 2 Small achieved a score of 31 on the Artificial Analysis Intelligence Index. This score ranks it competitively alongside other prominent models like Gemini 3.8 Flash. For perspective, leading models such as Gemini 4 Carbon and Claude Opus 5.5 boast scores of 60 and 59 respectively.

Upcoming Releases and Industry Trends

Thinking Machines has announced plans to release the weights of Inkling 2 Small next week, a standard practice intended to ensure thorough safety and security testing before public availability. This cautious approach aligns with wider industry trends, especially in light of increasing security concerns regarding AI technologies.

The timing of Inkling 2 Small’s launch coincides with that of Mistral’s Le Chonk model, which has quickly gained traction for its superior coding abilities. With these developments, Thinking Machines is poised to become a leading name in the Western open-weight AI arena.

Strengths and Areas for Improvement

Although Inkling 2 Small shows remarkable growth with a 41% improvement in automation and a 20% increase in coding proficiency over its predecessor, reviewers at Artificial Analysis have pointed out its slower processing speeds and tendency to produce verbose outputs. The model is priced at $0.30 per million input tokens, slightly higher than the $0.24 average for its category, which could lead to hesitation among potential users.

Accuracy Enhancements

Recent evaluations have demonstrated noteworthy advancements in accuracy, with a reduction in the hallucination rate from 61% to 27%. Additionally, its overall accuracy rose from 32% to 34%, with a drop in the attempt rate from 77% to 53%, indicating improved alignment with expected outputs.

With the unveiling of Inkling 2 Small, Thinking Machines is highlighting its commitment to advancing the AI field while navigating the competitive landscape shaped predominantly by various Chinese firms. In this ongoing race for innovation, the ramifications of this release will be watched closely by stakeholders across the industry.