Kimi K3.1 Pro beats GPT-6 Astra on Terminal-Bench 4.0
Moonshot's Kimi K3.1 Pro scores 65.7% on Terminal-Bench 4.0 versus GPT-6 Astra's 58.2%, at roughly one-seventh the price.
A $50 billion Chinese AI startup just released a model that beats GPT-6 Astra on coding benchmarks. Most Americans have never heard its name. Moonshot AI released Kimi K3.1 Pro today October 24, 2026. It scored 65.7% on Terminal-Bench 4.0, the benchmark that tests coding agents on real-world tasks. GPT-6 Astra scored 58.2%. Claude Opus 5.5 scored 66.4%. Kimi K3.1 Pro sits between the two most powerful AI models on earth and costs a fraction of both. $2 per million input tokens. $10 per million output tokens. Claude Opus 5.5 charges $15/$75. GPT-6 Astra charges more. Kimi is 7-8x cheaper on input and 7.5x cheaper on output. The model is built on Kimi K3 the largest open-weight model ever released at 2.8 trillion parameters. Full model weights for K3.1 Pro will be open-sourced by November 4. Bloomberg called it “Moonshot’s first major upgrade since its own DeepSeek moment.” And most people in the West have never typed “http://kimi.ai” into a browser. Here’s everything about the AI model the US tech giants don’t want you to discover 🧵