Moonshot AI cuts prices as its new model closes the gap with top US rivals

Moonshot AI has launched Kimi K3.1 Pro, a significant update to its flagship reasoning model that immediately repositioned the Chinese firm in the global market. The release places the new model at number four on the Artificial Analysis Intelligence Index, scoring 55 points. This marks a substantial leap from its predecessor, Kimi K3, which scored 44 points, narrowing the performance gap with the leading proprietary models from the United States.

The primary competitive advantage of the new release lies in its price-performance ratio. While the model maintains the same architectural scale as its predecessor, Moonshot has reduced API pricing by 33 percent. Input tokens now cost $2 per million, down from $3, while output tokens are priced at $10 per million, a drop from $15. Cache hits have also seen a reduction, falling to $0.20 from $0.30. Despite the price cut, the effective cost per task on the Intelligence Index remains nearly identical to the previous version at $1.98, largely because the new model generates 14 percent more output tokens to complete tasks.

Performance improvements are most visible in agentic tasks, where the model demonstrates enhanced autonomy. On the GDPval-AA benchmark, Kimi K3.1 Pro achieves an Elo score of 1,846, a 312-point increase over the previous iteration. This places it just behind Anthropic’s Claude Opus 5.5, which scores 1,866, and Muse Wonder Full, which records 1,857. In coding evaluations, the model shows dramatic progress on Terminal-Bench 4.0, reaching 65.7 percent accuracy. This is a massive jump from the 13 percent recorded by Kimi K3 and slightly trails Claude Opus 5.5, which sits at 66.4 percent.

The company also highlights improvements in reliability and specific professional domains. Internal testing on the Kimi Hallu Bench Hard shows a sharp decline in hallucination rates, dropping from 23.4 percent to 8.3 percent. While this is still higher than the 5.5 percent rate for GPT-6 Astra, Moonshot notes this benchmark serves as a stress test rather than a measure of general error rates. In finance, the model gained 7.2 points in modeling and 11.6 points in research, while also topping all tested models on the OfficeQA Pro v2 benchmark for general office tasks.

A notable expansion of the model's capabilities is its integration with 3D modeling and game development environments. Kimi K3.1 Pro can directly interact with Unity, Unreal Engine, and Godot, handling both minor adjustments and full project builds. On the BenchCAD benchmark, it scores 92.7 percent when using tools and 72.8 percent without them, figures that are close to those of Claude Opus 5.5. Moonshot has also introduced a "3D Make" feature on its website, allowing users to preview, edit, and export 3D models generated by the system.

The model retains the same parameter count as its predecessor, with 2.8 trillion total parameters and 104 billion active parameters. It supports a context window of one million tokens and processes both text and image inputs. According to Artificial Analysis, the model sits on the Pareto frontier, meaning no other model with a comparable size achieves a higher score on the intelligence index. This efficiency suggests that scale alone is no longer the sole driver of performance in the current market landscape.

A key strategic move for Moonshot is the planned release of the model’s weights by November 4, 2026. If executed as scheduled, Kimi K3.1 Pro would become the highest-scoring open-weights model available. This would reduce the performance gap between open and proprietary models from 12 points to just three, assuming no other major releases occur in the interim. This move could pressure competitors to either improve their open-source offerings or differentiate their proprietary services further.

Moonshot has also addressed cybersecurity concerns, noting that vulnerability research capabilities grew as post-training compute increased. To manage this, the company stripped cyber-related data from the training phases for the open-weights release. On an internal benchmark reproducing recent vulnerabilities in software like the Linux kernel, the uncleaned version solved 90.5 percent of tasks, while the cleaned version solved 81.0 percent. This is higher than the 33.3 percent scored by the previous Kimi K3 but lower than DeepSeek-V4.1-Flash, which reached 85.7 percent.

The company argues that no open-weights model should hold the absolute frontier of offensive cyber capabilities, but it aims to gradually increase what open models can do for defensive purposes, such as auditing legacy systems. This stance reflects a broader industry debate about the balance between openness and security. By releasing a model that is competitive with top US rivals at a lower price point, Moonshot is signaling that the cost of high-performance AI is dropping, potentially intensifying competition in the enterprise and developer markets.

Both the standard Kimi K3.1 Pro and a faster variant called K3.1 Pro Highspeed are available now. The high-speed version runs two to three times faster according to Moonshot, offering a trade-off between latency and cost. The models are accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. As the market continues to evolve, the entry of a highly capable, cost-effective model from a Chinese provider adds another layer of complexity for businesses choosing their AI infrastructure.