Moonshot AI Launches Kimi K3.1 Pro with 1M-Token Context
Original: Kimi K3.1 Pro and K3.1 Pro Highspeed
Why This Matters
Open-weight models approaching top proprietary benchmarks raises the competitive ceiling.
Moonshot AI released Kimi K3.1 Pro, its most capable 3T-class open-weight model, featuring a 1-million-token context window and a faster Highspeed variant. Full weights drop Nov 4.
Moonshot AI has launched Kimi K3.1 Pro, the first model in its K3.1 series, built on post-training of the existing Kimi K3 base with significantly more reinforcement learning compute. The company says it optimized its entire stack — RL framework, long-context inference, sandbox infrastructure, and rollout scheduling — to extract stable gains from each additional compute increment across hundreds of downstream domains.
On Terminal-Bench 4.0, which tests coding agents on resource-constrained terminal tasks, K3.1 Pro scores 65.7%, beating GPT-6 Astra (58.2%) and landing just below Claude Opus 5.5 (66.4%). On Moonshot's internal Kimi Code Bench 3.0 Long-horizon benchmark, it scores 61.2%, up from K3's 55.95%.
The model targets coding, 3D creation, financial research, and agentic workflows. It works across game engines including Unity, Unreal Engine, and Godot, pairing them with web search and 3D generation. A companion release, Kimi K3.1 Pro Highspeed, runs 2–3× faster for latency-sensitive use cases. Both are live now on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Full open weights are scheduled for November 4.