Kimi K3.1 Pro (runware)

Moonshot AI launched Kimi K3.1 Pro on October 24, 2026, alongside Kimi K3.1 Pro Highspeed, which the company says runs two to three times faster with identical model capabilities. The base model is described as Moonshot's most capable 3T-class open-weight release to date, carrying a 1-million-token context window. Both models are exposed across the Kimi family of products. Availability spans kimi.ai and the iOS, Android, and HarmonyOS mobile apps, the Kimi Work desktop app version 3.1.0 or later on Windows and Apple silicon Macs, and Kimi Code via the /model command. On the Kimi API the models appear as kimi-k3.1-pro and kimi-k3.1-pro-highspeed, with Zero Data Retention available to API customers, and enterprise administrators can enable K3.1 Pro for their organizations, off by default at launch.

Moonshot AI introduced Kimi K3.1 Pro as the first model in its K3.1 series, describing it as its most capable 3T-class open-weight model with a 1-million-token context window and substantial gains in coding, 3D creation, and agentic tasks. It is a post-trained version of Kimi K3 built with substantially larger post-training compute and large-scale joint reinforcement learning across hundreds of downstream domains. The release required stack-level optimization across the RL framework, long-context inference, training efficiency, sandbox infrastructure, rollout scheduling, and environment quality checks. Moonshot reports K3.1 Pro scoring 65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra (58.2%) and just behind Claude Opus 5.5 (66.4%), and 61.2% on its internal Kimi Code Bench 3.0 Long-horizon, up from K3's 55.95%. A Highspeed sibling variant claims 2-3x the speed with identical capabilities.

Plotly tested the exact Kimi K3.1 Pro model on its data analytics exam, where it passed 42 of 43 questions (98%) versus 35 for Kimi K3, tying Grok 4.7 for third place on a 34-model leaderboard behind GPT-6.1 Sol and GPT-6 Astra. Plotly called it the most accurate open-weight model it had tested, ahead of MiniMax M3.1 Flash's 41, with full weights released on November 2. The gains came from initiative rather than basic correctness: K3.1 Pro found all five hidden patterns and caught all three data quality problems that K3 missed, at about 7.5 model calls per question and roughly 73 tool calls on each hidden-pattern question versus about 18 for GPT-6.1 Sol. A full exam run cost Plotly $13.47 with a 2.5-minute median time per question, up from K3's $5.01 and 43 seconds.

Moonshot AI introduced Kimi K3.1 Pro, the first model in its K3.1 series and its most capable 3T-class open-weight model so far, with a 1-million-token context window. The post-trained successor to Kimi K3 was scaled with large joint reinforcement learning across hundreds of downstream domains, improving the RL framework, long-context inference, training efficiency, sandbox infrastructure, and rollout scheduling. Moonshot reports Kimi K3.1 Pro scoring 65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra at 58.2% and just behind Claude Opus 5.5 at 66.4%, while its internal Kimi Code Bench 3.0 Long-horizon score rose to 61.2% from K3's 55.95%. The model is offered on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, with a Kimi K3.1 Pro Highspeed variant delivering roughly 2-3x faster output.