Moonshot AI Narrows Gap With US Leaders as Kimi K3.1 Pro Debuts

Moonshot AI has launched Kimi K3.1 Pro, a major update to its reasoning model that immediately positions the Chinese firm as a primary competitor to leading US providers. The new release scores 55 points on the Artificial Analysis Intelligence Index, securing the fourth-highest rank globally. This places it just three points behind Anthropic’s Claude Opus 5.5, which leads the proprietary sector with 58 points. The launch marks a significant shift in the competitive landscape, demonstrating that Chinese AI development is rapidly closing the performance gap with its Western counterparts.

The most notable improvements are found in agentic tasks and coding capabilities. On the GDPval-AA benchmark, K3.1 Pro reaches an Elo score of 1,846, an increase of 312 points over its predecessor. In terminal-based coding tests, the model achieves a 65.7 percent success rate on Terminal-Bench 4.0, narrowly trailing Claude Opus 5.5 and significantly outperforming OpenAI’s GPT-6 Astra. Additionally, Moonshot reports a sharp decline in hallucination rates, dropping from 23.4 percent to 8.3 percent on internal stress tests, though GPT-6 Astra still holds a lower error rate of 5.5 percent.

The model also demonstrates enhanced capabilities in 3D modeling and software integration. It can interact directly with game engines such as Unity and Unreal Engine, scoring 92.7 percent on the BenchCAD benchmark when using tools. Moonshot has introduced a new '3D Make' feature on its platform, allowing users to preview and export 3D models generated by the AI. These advancements suggest the model is becoming a versatile tool for developers working on complex, multi-faceted projects that require both code generation and spatial reasoning.

Despite the performance gains, Moonshot has maintained the same architectural dimensions as the previous version, utilizing 2.8 trillion total parameters with 104 billion active during inference. The context window remains at one million tokens. However, the company has reduced API pricing by 33 percent, lowering input costs to $2 per million tokens and output costs to $10. This price reduction aims to make high-performance reasoning more accessible for enterprise applications without compromising on computational scale.

A key aspect of the release is Moonshot’s approach to cybersecurity capabilities. The company noted that vulnerability detection abilities improved naturally as post-training scaled. For the open-weights version, planned for release by November 4, 2026, Moonshot removed specific cyber-related data from the training corpus to manage offensive capabilities. This results in a slight performance drop on internal security benchmarks compared to the uncleaned version, balancing the model’s ability to find bugs with the need to limit its potential for offensive use.

If the open-weights version is released as scheduled, Kimi K3.1 Pro would become the highest-scoring open model available. This would reduce the performance gap between open and proprietary models from 12 points to just three, fundamentally changing the economics of AI deployment. For California-based tech firms and developers, this shift suggests that high-end AI capabilities will become more cost-effective and accessible, potentially altering the strategic advantage currently held by proprietary systems.

The immediate availability of K3.1 Pro and its faster variant, K3.1 Pro Highspeed, through various API endpoints allows for rapid integration. As the industry watches the upcoming open-weights release, the focus will remain on whether this model can sustain its performance lead in specialized domains like finance and security. The move signals a continued trend toward democratizing advanced AI tools while intensifying competition at the frontier of large language models.