New Kimi model cuts API costs by 33 percent to rank fourth

Moonshot AI released Kimi K3.1 Pro on October 23, 2026, a reasoning model that now ranks fourth on the Artificial Analysis Intelligence Index. The system scored 55 points, an 11-point gain over the previous Kimi K3 version. It sits three points behind the leading proprietary systems, marking the closest approach by a Chinese model to the top tier.

The score places K3.1 Pro behind Anthropic's Claude Opus 5.5, which holds 58 points, and Claude Sonnet 5.5 and Muse Wonder Full, both at 56. The update comes with a 33 percent reduction in API pricing. Input tokens now cost $2 per million, output tokens cost $10 per million, and cache hits cost $0.20 per million.

Performance gains were most pronounced in agentic tasks. On the GDPval-AA benchmark, K3.1 Pro reached 1,846 Elo, a 312-point increase from the K3 model. It trails only Claude Opus 5.5 at 1,866 Elo and Muse Wonder Full at 1,857 Elo. In Terminal-Bench 4.0 testing, the new model achieved a 65.7 percent success rate. This surpasses GPT-6 Astra, which scored 58.2 percent, and falls slightly short of Claude Opus 5.5 at 66.4 percent. The previous K3 model managed only 13 percent on this test.

Moonshot reports that claim-level hallucinations dropped from 23.4 percent in K3 to 8.3 percent in K3.1 Pro. The company tested this on its internal Kimi Hallu Bench Hard, describing it as a stress test for error rates. GPT-6 Astra still leads this specific benchmark at 5.5 percent. In finance-related tasks, the model improved by 7.2 points on Financial Modeling and 11.6 points on Financial Research. Moonshot states that K3.1 Pro outperforms all tested models on OfficeQA Pro v2.

The model maintains the same architecture as its predecessor, featuring 2.8 trillion total parameters with 104 billion active. It supports a one-million-token context window and processes text and images. Despite generating 14 percent more output tokens, the cost per Intelligence Index task remains nearly identical to the previous version at $1.98, compared to $2.00 for K3.

For developers, Moonshot offers a Highspeed variant that the company says runs two to three times faster. Both versions are accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. The company also introduced a 3D Make feature on Kimi.com for previewing and exporting models. K3.1 Pro integrates with game engines like Unity and Unreal Engine, and it can write scripts for Blender and CAD software without additional connectors.

On the BenchCAD benchmark, the model scored 92.7 percent when using tools. This compares to 96.2 percent for Claude Opus 5.5 and 95.9 percent for GPT-6 Astra. Without tools, K3.1 Pro reached 72.8 percent, slightly below Claude Opus 5.5 at 73.0 percent.

Moonshot plans to publish the model weights by November 4, 2026. If released as scheduled, K3.1 Pro would become the highest-scoring open-weights model available. This would shrink the gap between open and proprietary systems from 12 points to 3 points. The company removed cybersecurity-related data from the training corpus for the open release to manage offensive capabilities.

On internal cyber benchmarks, the uncleaned model solved 90.5 percent of tasks, while the cleaned version solved 81.0 percent. DeepSeek-V4.1-Flash scored 85.7 percent on the same test, compared to 33.3 percent for the original K3 model. Moonshot aims to raise the capabilities of open models gradually to support defensive auditing work.