Kimi K3.1 Pro, the strongest open model, now on Berget AI
Developer Relations
On this page
Last week, Moonshot AI published the weights to Kimi K3.1 Pro, the strongest open-weight model to date. Starting today, you can use it on Berget AI, on our own hardware in Sweden, under EU jurisdiction—your data never leaves Europe.
Kimi K3.1 Pro replaces Kimi K3, at €3/Mtok (input) and a reduced €10/Mtok (output), against €3 and €15 for K3.
What the benchmarks say
Artificial Analysis ranks K3.1 Pro first among open-weights models on its current Intelligence Index. Vals AI scores it as the top open-weights model on the Vals Index at 62.52%, ahead of DeepSeek V4.1-Flash-1017 at 61.26%. And its 67% on Terminal-Bench 4.0 is the highest score Artificial Analysis has recorded for any model.
Kimi K3.1 Pro excels at agentic coding. On Terminal-Bench 4.0, where an agent gets a terminal and a real task, K3.1 Pro scores 67%, up from 13% for K3 and ahead of Claude Sonnet 5.5 at 64%. At launch it entered the Intelligence Index fourth of 215 models, behind Claude Opus 5.5.
Image credit: Artificial Analysis
On the current index K3.1 Pro scores 46 against 58 for Claude Opus 5.5, a 12-point gap to the closed leaders.
Run Kimi K3.1 Pro today
You can use Kimi K3.1 Pro today using any of our Berget Code plans, from the Playground, or directly from our OpenAI-compatible API.
curl https://api.berget.ai/v1/chat/completions \ -H "Authorization: Bearer $BERGET_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "moonshotai/Kimi-K3.1-Pro", "messages": [{"role": "user", "content": "Your prompt here"}] }'