--- title: "Kimi K2.6 API hosted in the EU" description: "Run Moonshot AI's Kimi K2.6 on inference.eu: a 1-trillion-parameter open agentic model with image input and 256K context, processed in EU data centres." canonical: "https://inference.eu/models/kimi-k2-6/" language: "en" --- Reasoning · Moonshot AI # Run Kimi K2.6 in the EU through an OpenAI-compatible API Kimi K2.6 is Moonshot AI's open-weight agentic model from April 2026: one trillion parameters with 32 billion active, native image understanding and a 256K-token context. On inference.eu you call it with the OpenAI or Anthropic request format, while your prompts and code are processed in EU data centres and never used for training. [Contact sales](https://inference.eu/contact/)[Example request](#example) 1. 011T · 32B 2. 02EU data centres 3. 03Modified MIT License ## What Kimi K2.6 is Kimi K2.6 is a mixture-of-experts model from Moonshot AI in Beijing, released with open weights on 20 April 2026. It has one trillion parameters in total, 32 billion active per token, 61 layers and a context window of 256K tokens. A 400-million-parameter vision encoder called MoonViT gives it native image input. Moonshot describes it as built for long-horizon coding, turning prompts and screenshots into working interfaces, autonomous execution and swarm-style orchestration, where one run splits work across many sub-agents. It has a thinking mode and an instant mode without reasoning. The licence is a modified MIT licence with an attribution clause for very large products. ## What it does well, and where it falls short K2.6 is strong at end-to-end coding across languages such as Rust, Go and Python, at front-end work from a sketch or screenshot, and at long agent runs that call many tools. The instant mode makes it usable for faster chat-style tasks too, so one model can cover both quick answers and deep agent work. With one trillion parameters it is a heavy model; for simple tasks a smaller one is more economical. Its context of 256K is shorter than the one-million-token windows of Kimi K3, GLM-5.3 or DeepSeek V4. Moonshot's newer Kimi K3 is more capable overall. As with other Chinese models, its training reflects its maker's choices on sensitive topics. ## Calling Kimi K2.6 through the API inference.eu offers the OpenAI and the Anthropic request formats. Set the base URL to the inference.eu API, use your inference.eu key and set the model to `kimi-k2.6`. Coding agents such as Claude Code, Cline or OpenHands work by changing those same settings. Thinking mode is the default. To use instant mode, pass `thinking: false` in the chat template options through `extra_body`. Moonshot recommends `temperature` 1.0 for thinking mode and 0.6 for instant mode, with `top_p` 0.95. Send images as `image_url` content parts. Tool calls and streaming work as usual, and the reasoning comes back separately from the answer. ## Data residency for Kimi K2.6 Kimi K2.6 comes from China, but on inference.eu its open weights run on infrastructure in the European Union. Your prompts, images, code and the model's responses are processed in EU data centres, are not sent to Moonshot AI, and are never used to train any model. Every customer receives a data processing agreement under Article 28 of the GDPR and our subprocessor list. Billing is in euros. The modified MIT licence allows commercial use; its only addition is that products with more than 100 million monthly active users or more than 20 million US dollars in monthly revenue must display "Kimi K2.6" prominently in their interface. ## Kimi K2.6 Maker Moonshot AI Made in China Type Reasoning Parameters 1T Active parameters 32B Context window 262,144 tokens Licence [Modified MIT License](https://huggingface.co/moonshotai/Kimi-K2.6/blob/main/LICENSE) Released April 2026 Model ID kimi-k2.6 Processing EU data centres Training on your data Never Source [huggingface.co](https://huggingface.co/moonshotai/Kimi-K2.6) kimi-k2.6 · Python ``` import os from openai import OpenAI client = OpenAI(base_url="https://api.inference.eu/v1", api_key=os.environ["INFERENCE_EU_API_KEY"]) response = client.chat.completions.create( model="kimi-k2.6", messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}], ) print(response.choices[0].message.content) ``` ## What teams use it for - 1 ### Front-end from screenshots Turn a design screenshot or a rough sketch into working HTML, CSS and components, then refine them in an agent loop. - 2 ### Long autonomous coding runs Let an agent carry a multi-file change through planning, implementation and testing with many tool calls. - 3 ### Parallel research agents Split a research or analysis task across several sub-agents and merge their findings into one report. ## Kimi K2.6: common questions The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly. Compliance depends on how you process data. On inference.eu, Kimi K2.6 requests are processed in EU data centres, not used for training and covered by a data processing agreement under Article 28 GDPR with a published subprocessor list. No data is sent to Moonshot AI or to China. On inference.eu the open weights run in EU data centres, so prompts and code never leave the EU for inference. What remains is the model's own training, which reflects its maker's choices, including on politically sensitive topics. For coding and agent work this rarely matters, but test it on your tasks. Kimi K2.6 supports 256K tokens of context, which is 262,144 tokens. That covers large parts of a repository or long agent histories. If you need one million tokens, Kimi K3, GLM-5.3 or DeepSeek V4 Flash offer longer windows. Yes. K2.6 uses a modified MIT licence that allows commercial use, modification and redistribution. The single modification: products with more than 100 million monthly active users or more than 20 million US dollars of monthly revenue must display Kimi K2.6 prominently in their user interface. Thinking mode is on by default. For instant mode, pass \`thinking: false\` inside \`chat\_template\_kwargs\` in the request's \`extra\_body\`. Moonshot recommends \`temperature\` 1.0 when thinking and 0.6 in instant mode, with \`top\_p\` 0.95 in both. Yes. Kimi K2.6 is natively multimodal and uses a 400-million-parameter vision encoder. It can read screenshots, designs, charts and photos together with text and answer in text. Send images in the standard OpenAI \`image\_url\` format. Kimi K3 is Moonshot's newer flagship, with 2.8 trillion parameters, 104 billion active, a one-million-token context and video understanding. K2.6 is smaller, with a 256K context and an instant mode. K3 is stronger; K2.6 is lighter and its licence is simpler for most companies. Yes. inference.eu offers an Anthropic-compatible endpoint, so Claude Code and other tools using the Anthropic Messages format can call Kimi K2.6. Change the base URL, your API key and the model name to \`kimi-k2.6\`. Set your OpenAI client's base URL to the inference.eu API, use your inference.eu key and change the model to \`kimi-k2.6\`. Messages, images, tools and streaming keep working. Use instant mode if you want answers without a reasoning phase. No. Moonshot AI published the weights but has no part in serving your requests on inference.eu. Your prompts and code are processed in EU data centres, used only for your response and never used to train Kimi or any other model. [Ask a question](https://inference.eu/contact/) ## Related [All models](https://inference.eu/models/#directory) - [Kimi K3Run Kimi K3, Moonshot AI's 2.8-trillion-parameter open-weight model with 1M context and image and video input, on inference.eu with processing in the EU.Reasoning](https://inference.eu/models/kimi-k3/) - [GLM-5.3Run Z.ai's GLM-5.3, an open-weight model built for complex coding and long agent tasks, on inference.eu with OpenAI and Anthropic APIs, processed in the EU.Reasoning](https://inference.eu/models/glm-5-3/) - [MiniMax M3Run MiniMax M3 on inference.eu: an open multimodal model with 1M context for coding and agent work, OpenAI-compatible and processed in EU data centres.Reasoning](https://inference.eu/models/minimax-m3/) - [Qwen3.8 27BRun Qwen3.8 27B on inference.eu: a dense open model with image and video input, adjustable reasoning and 256K context, OpenAI-compatible, processed in the EU.Reasoning](https://inference.eu/models/qwen3-8-27b/) ## Move your AI workloads to EU infrastructure Tell us which models you use and roughly how much traffic you have. We reply within one working day with pricing and next steps. - Reply within one working day - Pricing for your volume - DPA ready to sign