Key Takeaways
- Moonshot AI has launched Kimi K3.1 Pro, a new open-weight AI model.
- K3.1 Pro scores higher than GPT-6 Astra and matches Claude Opus 5.5 in coding and knowledge tasks.
- The model is cheaper to use and faster than its predecessor.
Moonshot AI has introduced Kimi K3.1 Pro, a new open-weight AI model that competes with leading proprietary models in coding and knowledge tasks. According to the company, K3.1 Pro matches or outperforms GPT-6 Astra and comes close to Anthropic’s Claude Opus 5.5 in various benchmarks.
In Moonshot’s own benchmarks, K3.1 Pro scores 65.7% on Terminal-Bench 4.0, surpassing GPT-6 Astra (58.2%) and falling just behind Claude Opus 5.5 (66.4%).
The new model is part of Moonshot’s K3.1 series and is described as the most capable 3T-class open-weight model so far. It features a 1-million-token context window and includes significant upgrades in coding, 3D creation, financial research, and other agent tasks.
K3.1 Pro is also cheaper to use than its predecessor through the API, making it an attractive option for developers and businesses looking to integrate AI capabilities into their operations.
Moonshot has also released a faster version of K3.1 Pro, called K3.1 Pro Highspeed, which runs two to three times quicker than the standard model. Both models are available on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
The full model weights will be released by November 4, allowing users to explore the full capabilities of K3.1 Pro.
K3.1 Pro has demonstrated superior performance in several tests, including SWE-Marathon, Humanity’s Last Exam Full with tools, OfficeQA Pro v2, and ZeroBench with tools. It also leads on BrowseComp and AutomationBench-AA.
However, K3.1 Pro still falls behind in certain areas, such as Terminal-Bench Science 0.1, ProgramBench, HLE-Diamond, and GPQA-Diamond. The model is intentionally weaker in cybersecurity tests like ExploitBench and ExploitGym, according to Moonshot.
Despite its limitations, K3.1 Pro offers a competitive alternative to top proprietary models, making it a valuable tool for developers and businesses seeking cost-effective AI solutions.
