Moonshot AI has officially released Kimi K3.1 Pro, its most powerful open-weight model to date, marking a significant leap in both parameter scale and intelligence for the open-source community. Built on a 2.8T total parameter MoE architecture (with ~104B active parameters), it supports a massive 1 million token context window and includes a new 'Highspeed' version offering 2-3x faster inference. Full weights are scheduled for release on November 4.
Benchmark data confirms Kimi K3.1 Pro's excellence in coding, financial analysis, and agentic tasks. It ranks fourth on the Artificial Analysis Intelligence Index, with Terminal-Bench scores surging from 13% to 67%. In code generation, it rivals Claude Opus-5 while costing 70% less. The company reports a 65% reduction in hallucinations compared to previous versions, positioning it as a top performer in complex multi-step reasoning tasks like OfficeQA Pro v2.
This release significantly impacts the developer ecosystem. As the first 3T-class open-weight model, Kimi K3.1 Pro is poised to surpass DeepSeek-v4.1-Flash as the leading open-source code model. Its aggressive pricing and high performance will accelerate enterprise adoption, narrowing the gap between open and closed models and reshaping the market's cost-performance balance.