Kimi K3.1 Pro Release and Core Features
Kimi K3.1 Pro is the first model in Kimi's new K3.1 series and the most capable 3T-class open-weight model to date, featuring a 1-million-token context window. Building on K3 with expanded post-training compute and optimized RL framework, long-context inference, and sandbox infrastructure, it delivers stable gains across coding, 3D creation, and financial research.
The Kimi K3.1 Pro Highspeed version offers 2-3 times the inference speed for responsive interactions. The model is available today on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, with full model weights releasing by November 4, 2026.
Multi-Domain Benchmark Performance and Real-World Applications
On coding benchmarks, K3.1 Pro scores 65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%). Its 61.2% on Kimi Code Bench 3.0 Long-horizon exceeds K3's 55.95%. The model handles quick fixes in real codebases, full project builds, and creates complete 3D experiences using engines such as Unity, Unreal Engine, and Godot.
On 3D modeling, it scores 92.7% on BenchCAD (with tools), approaching Claude Opus 5.5 (96.2%) and GPT-6 Astra (95.9%). The model infers shapes, proportions, and spatial relationships from text, sketches, or images to build printable 3D models, editing asset libraries via Blender and CAD scripts.
In financial research, it scores 64.4% on OfficeQA Pro v2, with significant improvements on Frontier Finance and internal benchmarks (Financial Modeling +7.2 points, Financial Research +11.6 points, Business Operation +4.2× K3). Targeted factual accuracy improvements reduce hallucination rate on Kimi Hallu Bench Hard from 23.4% (K3) to 8.3%, enabling auditable research.
Open-Weight Principles and Limitations
Through data cleaning and RL controls, K3.1 Pro's pass rate on xdev-bench-hard drops from 90.5% to 81%, while general coding capability remains stronger than its predecessor. However, its offensive cyber capability is below that of proprietary models publicly available for at least three months, by design. Full weights release November 4, 2026 for organizational enablement.