Moonshot AI releases Kimi K3.1 Pro to challenge top global models
The new reasoning model achieves a significant leap in performance benchmarks while reducing API costs for developers.
By Hiroshi Gupta ·

Moonshot AI has officially launched Kimi K3.1 Pro, a post-trained reasoning model that immediately secures the fourth position on the Artificial Analysis Intelligence Index. With a score of 55, the new release narrows the performance gap with leading proprietary systems. It now sits just three points behind Anthropic's Claude Opus 5.5, which leads the index with 58 points. This rapid advancement positions the Chinese AI firm as a more direct competitor in the high-end market.
The model demonstrates substantial improvements in agentic tasks and coding capabilities. On the GDPval-AA benchmark, Kimi K3.1 Pro achieves an Elo rating of 1,846. This represents a 312-point increase over its predecessor, Kimi K3. The score places it slightly behind Claude Opus 5.5 and Muse Wonder Full. In terminal-based evaluations, the new model reports a 65.7 percent success rate on Terminal-Bench 4.0. This result surpasses GPT-6 Astra and closely trails the current market leader.
Moonshot also reports a significant reduction in hallucination rates for the new release. Internal testing on the Kimi Hallu Bench Hard shows claim-level hallucinations dropping from 23.4 percent to 8.3 percent. The company notes this serves as a stress test rather than a measure of general error rates. GPT-6 Astra maintains a slight edge with a 5.5 percent rate on the same benchmark. Additionally, the model shows gains in financial modeling and research tasks, topping all tested models on OfficeQA Pro v2 for general office work.
A major focus of the update is 3D modeling and integration with game engines. The model can work directly with Unity, Unreal Engine, and Godot for both quick fixes and full project builds. It scores 92.7 percent on the BenchCAD benchmark when using tools. Without tools, the score is 72.8 percent, which nearly matches Claude Opus 5.5. Moonshot launched a feature called 3D Make on its website to allow users to preview, edit, and export these models. The system is particularly strong at parametric modeling of mechanical parts and architectural designs.
Despite the performance gains, the underlying architecture remains consistent with the previous version. Kimi K3.1 Pro retains 2.8 trillion total parameters, with 104 billion active parameters. It supports a context window of one million tokens and processes both text and images. The company has reduced API prices by 33 percent to improve accessibility. Input costs are now $2 per million tokens, down from $3. Output costs stand at $10, reduced from $15, while cache hits cost $0.20.
Moonshot plans to release the model weights by November 4, 2026. If this timeline holds, K3.1 Pro would become the highest-scoring open-weights model available. This release would shrink the gap between open and proprietary models from 12 points to just three. The company also offers a variant called K3.1 Pro Highspeed. This version runs two to three times faster and is currently available through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
The technical documentation includes a detailed section on cybersecurity measures. Moonshot observed that vulnerability research capabilities grew as post-training scaled up. To address this, the company stripped cyber-related data from both training phases for the open-weights release. On the internal xdev-bench-hard test, the uncleaned version solved 90.5 percent of tasks. After data cleaning, the model solves 81.0 percent. DeepSeek-V4.1-Flash reaches 85.7 percent on the same test, while the older Kimi K3 scored 33.3 percent.
The company aims to balance offensive capabilities with defensive utility. Moonshot states that no open-weights model should push the absolute frontier of offensive cyber capabilities. However, it intends to gradually raise what open models can do to support tasks like auditing legacy systems. Anthropic previously noted that even models scoring lower on the index can write exploits at high levels. This approach reflects a broader industry effort to manage the risks associated with releasing powerful AI models publicly.
The launch of Kimi K3.1 Pro marks a significant shift in the competitive landscape. By combining improved reasoning with lower costs, Moonshot challenges the dominance of established players. The upcoming release of open weights could further democratize access to top-tier AI performance. Readers in New York and New Jersey should note the growing influence of Chinese AI firms in global markets. The coming months will reveal how other vendors respond to these advancements.






