Moonshot AI's Kimi K3.1 Pro Narrows Gap With Top US Rivals

For developers and businesses relying on large language models, the cost of accessing top-tier intelligence is dropping. Moonshot AI has just released Kimi K3.1 Pro, a model that scores highly on major intelligence indices while cutting API prices by a third. This release marks a significant shift in the competitive landscape, as a Chinese-made model now sits just three points behind the leading proprietary systems from American companies.

The new model, a post-trained evolution of the previous Kimi K3, secured a score of 55 on the Artificial Analysis Intelligence Index. This result places it in fourth position globally, trailing only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. If Moonshot releases the model's weights by early next month as planned, K3.1 Pro would become the highest-scoring open-weights model available. That move would reduce the performance gap between open and proprietary systems from twelve points to just three, assuming no other major releases occur in the interim.

Performance gains are most visible in agentic tasks and coding. On the GDPval-AA benchmark, the model achieved an Elo rating of 1,846, a substantial leap from its predecessor. It also posted a 65.7 percent score on Terminal-Bench 4.0, a metric where the older K3 model managed only 13 percent. Moonshot reports that hallucination rates have dropped sharply, with claim-level errors falling from 23.4 percent to 8.3 percent on its internal stress tests, though rival GPT-6 Astra still holds a lower error rate.

The model also demonstrates strong capabilities in 3D modeling and software development. It can interact directly with game engines such as Unity and Unreal Engine, and it scored 92.7 percent on the BenchCAD benchmark when using tools. This performance is close to that of Claude Opus 5.5 and GPT-6 Astra. Moonshot has launched a feature called "3D Make" that allows users to preview and export models, highlighting the model’s strength in parametric modeling for mechanical and architectural design.

Despite these improvements, the model maintains the same architectural size as its predecessor, with 2.8 trillion total parameters. The context window remains at one million tokens. However, the price per million input tokens has dropped to $2, and output costs have fallen to $10. This makes the model more accessible for high-volume applications without a reduction in core capacity.

Moonshot has also addressed cybersecurity concerns by removing specific cyber-related data from the training process for the open-weights version. This deliberate limitation reduces its offensive capabilities on internal vulnerability tests, a trade-off the company argues is necessary for open models. The company notes that while it wants to raise the bar for open models, they should not yet push the absolute frontier of offensive cyber capabilities, balancing the need for defensive auditing tools against the risks of widespread access.

Both the standard K3.1 Pro and a faster "Highspeed" variant are available now through Kimi’s various platforms. The release signals an intensifying competition in the AI sector, where performance and price are increasingly converging across national borders.