AI inference costs drop 33 percent as new Chinese model challenges top rivals

Developers and businesses relying on advanced artificial intelligence services will see a notable reduction in operating costs starting this week. Moonshot AI has introduced a new reasoning model that significantly lowers the price per token while simultaneously improving performance in specialized technical tasks. This shift aims to make high-end AI capabilities more accessible to a broader range of users without sacrificing quality.

The primary change for ordinary users and companies is the 33 percent decrease in API pricing. Input costs have fallen from three dollars to two dollars per million tokens. Output costs have dropped from fifteen dollars to ten dollars per million tokens. Additionally, the cost for cache hits has decreased from thirty cents to twenty cents. These price adjustments apply to the new Kimi K3.1 Pro model, which is available immediately through various platforms including Kimi.ai and the Kimi API.

Performance metrics indicate that the new model has closed the gap with leading proprietary systems. It currently holds the fourth position on the Artificial Analysis Intelligence Index with a score of 55 points. This represents an 11 point improvement over its predecessor, Kimi K3. The model trails only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. If Moonshot releases the model weights as scheduled for November 4, 2026, it would become the highest-scoring open-weights model. This would narrow the performance gap between open and proprietary models from twelve points to just three.

The most significant technical improvements are observed in agentic tasks and coding. On the GDPval-AA benchmark, the new model achieves a score of 1,846 Elo. This is 312 points higher than the previous version. It places the model just behind Claude Opus 5.5 and Muse Wonder Full. In terminal-based coding tests, the model scores 65.7 percent on Terminal-Bench 4.0. This surpasses GPT-6 Astra and comes within a narrow margin of Claude Opus 5.5. The previous version of Kimi scored only 13 percent on this specific test.

The model also demonstrates enhanced capabilities in 3D modeling and financial analysis. It can interact directly with game engines such as Unity, Unreal Engine, and Godot. On the BenchCAD benchmark, it achieves a 92.7 percent score when using tools. This is close to the scores recorded by Claude Opus 5.5 and GPT-6 Astra. The company also reports improvements in financial modeling and research, with gains of 7.2 and 11.6 points respectively. For general office tasks, the model reportedly outperforms all other tested systems on the OfficeQA Pro v2 benchmark.

Despite the performance gains, the underlying architecture remains consistent in size. The model retains the 2.8 trillion total parameters and 104 billion active parameters of its predecessor. It supports a context window of one million tokens and processes both text and images. According to Artificial Analysis, the model sits at the top of the Pareto frontier. No other model with a comparable parameter count currently scores higher on the Intelligence Index. A faster variant called K3.1 Pro Highspeed is also available, which runs two to three times quicker.

Moonshot AI has taken specific steps to manage the model's cybersecurity capabilities for its open-weights release. The company removed cyber-related data from both training phases to balance offensive and defensive potential. On an internal benchmark that reproduces known vulnerabilities in software like the Linux kernel, the cleaned model solves 81.0 percent of tasks. The uncleaned version solved 90.5 percent. For comparison, DeepSeek-V4.1-Flash reaches 85.7 percent on the same test. The company states that open-weights models should not push the absolute frontier of offensive cyber capabilities but should gradually improve for defensive auditing purposes.

The release marks a significant moment in the competition between open and proprietary AI systems. By lowering costs and improving specific technical skills, Moonshot AI positions its model as a viable alternative to top-tier closed systems. The planned release of model weights in November 2026 could further democratize access to high-performance AI. This development may influence pricing strategies across the industry as competitors adjust to the new market dynamics.

As the AI landscape continues to evolve, the focus is shifting toward efficiency and specialized capabilities. The new model's ability to handle complex 3D and coding tasks at a lower price point may drive broader adoption in industries ranging from software development to architectural design. The reduction in hallucination rates also adds to the model's reliability for professional applications. Users can now access these features through existing Kimi platforms, with the full open-weights release expected later this year.