Moonshot AI's Kimi K3.1 Pro Narrows Gap With Claude Opus 5.5
The new Chinese model ranks fourth globally while cutting API costs by a third and releasing open weights by November.
By Eduardo Salazar ·
Developers and businesses relying on large language models now have a more cost-effective option that performs nearly at the top tier. Moonshot AI has launched Kimi K3.1 Pro, a reasoning model that scores 55 points on the Artificial Analysis Intelligence Index. This places it in fourth position, just three points behind the leading proprietary systems. The release marks a significant shift in the competitive landscape, as the new model offers high performance at a reduced price point.
The primary change for users is a 33 percent drop in API pricing, with input tokens now costing $2 per million and output tokens at $10 per million. Despite the price cut, the model retains the same 2.8 trillion total parameter structure as its predecessor, Kimi K3. Moonshot plans to release the full model weights by November 4, 2026. If this timeline holds, K3.1 Pro would become the highest-scoring open-weights model available, shrinking the performance gap between open and proprietary models from 12 points to just three.
Performance gains are most evident in agentic tasks and coding. On the GDPval-AA benchmark, the model achieves 1,846 Elo, a 312-point increase over K3. It trails only Claude Opus 5.5 and Muse Wonder Full in this category. In terminal-based coding tests, K3.1 Pro records a 65.7 percent success rate on Terminal-Bench 4.0, far surpassing the 13 percent achieved by its predecessor. The company also reports a sharp decline in hallucinations, with claim-level errors dropping from 23.4 percent to 8.3 percent on its internal stress test, though GPT-6 Astra maintains a lower error rate of 5.5 percent.
The model demonstrates specific strengths in 3D modeling and finance. It can interact directly with game engines such as Unity and Unreal Engine, scoring 92.7 percent on the BenchCAD benchmark when using tools. Moonshot states the system is particularly effective at parametric modeling of mechanical parts and architectural designs. It also writes and runs scripts for Blender and CAD software without requiring additional connectors. A new feature called '3D Make' on Kimi.com allows users to preview and export these models directly.
To manage cybersecurity risks in the upcoming open-weights release, Moonshot removed cyber-related data from both training phases. On an internal benchmark replicating recent software vulnerabilities, the uncleaned version solved 90.5 percent of tasks, while the cleaned version solved 81.0 percent. This is slightly below DeepSeek-V4.1-Flash, which reached 85.7 percent. The company argues that while open models should not define the absolute frontier of offensive capabilities, they serve important defensive roles such as auditing legacy systems.
A faster variant, K3.1 Pro Highspeed, is also available, running two to three times quicker than the standard version. Both models are accessible through Kimi.ai, Kimi Work, Kimi Code, and the API. The standard model handles up to one million tokens of context and supports both text and image inputs. According to Artificial Analysis, it sits at the top of the Pareto frontier, meaning no model with a comparable parameter count scores higher on the Intelligence Index. Each task on the index costs approximately $1.98, nearly identical to the previous version due to increased output token generation.
Source: The Decoder
