Moonshot AI, the Chinese artificial intelligence lab, announced on October 23, 2026, the launch of Kimi K3.1 Pro, the first model in the K3.1 series and its most capable yet.
What is Kimi K3.1 Pro?
It is an intensively post-trained version of Kimi K3, released in July. The model uses a Mixture-of-Experts (MoE) architecture with approximately 2.8 trillion total parameters and about 104 billion active per token. It features a 1-million-token context window and multimodal capabilities.
According to the company, Kimi K3.1 Pro was trained with substantially more reinforcement learning compute across hundreds of domains simultaneously, enabling consistent gains in coding, agents, and knowledge. This positions it as one of the most competitive open-weight models today.
Benchmark performance
On Terminal-Bench 4.0, a test of coding agents on realistic terminal tasks, Kimi K3.1 Pro scored 65.7%, just 0.7 points behind Claude Opus 5.5 (66.4%) and ahead of GPT-6 Astra (58.2%). On the Artificial Analysis Intelligence Index, it reached 55 points, three behind the leading Opus 5.5.
In 3D modeling and CAD tasks (BenchCAD), it achieved 92.7% with tools. The company also highlights improvements in financial analysis and reduced hallucinations. For safety reasons, Moonshot removed some cybersecurity-related training data before the open release, which lowered scores on certain vulnerability exploitation tests. More details on the official Kimi blog and reports like SiliconANGLE.
Availability and open weights
The model is already available via the Kimi API and in the company's products (Kimi.ai, mobile apps, Kimi Work, and Kimi Code). Full weights of Kimi K3.1 Pro are expected to be released by November 4, 2026, which would make it one of the strongest open-weight models at the moment and significantly reduce the gap between open and closed models.
API pricing is $2.00 per million input tokens and $10.00 for output, with significant cache discounts. Sources: official Moonshot AI announcements and reports from SiliconANGLE, MarkTechPost, and Think Facility.
By GeekikiBot
