Kimi K2.6
Moonshot's April 2026 1T-parameter multimodal coding and agent model: 32B active, 256K context, native INT4, modified MIT.
In short. Kimi K2.6 is a very large AI model from Moonshot AI for long coding tasks and for agents that run many sub-agents. It also reads images and video. The license is the MIT license with one extra sentence that only affects very large products. It needs a server-class GPU node, and Moonshot has since released K2.7 Code and K3.
Summary
Moonshot AI released and open-sourced Kimi K2.6 on 20 April 2026 as its strongest coding model at the time. It has the same architecture as K2.5: a mixture of experts with about 1 trillion parameters and 32 billion active per token, 61 layers (one of them dense), 384 experts with 8 chosen per token plus one shared expert, multi-head latent attention, a 256K context and a 400-million-parameter MoonViT vision encoder for images and video. The weights are native INT4 and take about 595 GB. Moonshot's focus is long-horizon coding, interfaces built from prompts and images, and Agent Swarm, which it says grows from K2.5's 100 sub-agents and 1,500 steps to 300 sub-agents and 4,000 coordinated steps.
In Moonshot's own table K2.6 scores 58.6 on SWE-Bench Pro (K2.5 50.7, GPT-5.4 57.7, Claude Opus 4.6 53.4, Gemini 3.1 Pro 54.2), 80.2 on SWE-Bench Verified, 66.7 on Terminal-Bench 2.0 with Terminus-2 in preserve-thinking mode (K2.5 50.8, Gemini 3.1 Pro 68.5), 83.2 on BrowseComp, 54.0 on Humanity's Last Exam with tools, 90.5 on GPQA-Diamond and 73.1 on OSWorld-Verified (K2.5 63.3). Moonshot's help center adds that in its tests the model coded continuously for 13 hours. All of these are self-reported results.
Moonshot recommends vLLM, SGLang or KTransformers with transformers 4.57.1 or later, and says the K2.5 deployment recipe can be reused. K2.6 has an instant mode (temperature 0.6) and a thinking mode (temperature 1.0), plus an optional preserve-thinking mode, off by default, that keeps reasoning across turns for coding agents. Video input works only through Moonshot's own API. Moonshot later built K2.7 Code on K2.6, and its help center says Agent Swarm is now powered by K3, so we list K2.6 as a legacy model. In the 30 days to 11 November 2026, Hugging Face counted about 222,000 downloads. The license is Modified MIT: only a commercial product with over 100 million monthly active users or over US$20 million of monthly revenue must display "Kimi K2.6" in its interface.
Good for
- Self-hosted coding agents and long tool-call sequences on a GPU server
- Work that mixes screenshots, design images or video with code
- Companies that want an almost-MIT license for a trillion-parameter model
- Comparing the K2.5, K2.6, K2.7 Code and K3 generations
Watch out for
- K2.7 Code is newer and beats K2.6 in Moonshot's own table, and K3 is Moonshot's newer generation
- It needs a server-class GPU node, since even the INT4 weights are about 595 GB
- Video input works only through Moonshot's own API for now
- All benchmark numbers come from Moonshot, and the comparison models are from early 2026
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Estimates for the K2.6 (INT4) variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.
| Variants | Parameters | BF16 | FP8 / INT8 | 4-bit (Q4) | Fits on |
|---|---|---|---|---|---|
| K2.6 (INT4) INT4 | 1.03T 32B active | 2.36 TB | 1.18 TB | 655 GB | GPU server |
Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.
The license in plain words
The standard MIT license with one added sentence. Anyone can use, modify and sell the model, and only very large products (over 100 million monthly active users or over US$20 million of monthly revenue) must show the model's name in their interface.
- Attribution or model name must be shown
Frequently asked questions
- What is Kimi K2.6?
- Kimi K2.6 is a very large AI model from Moonshot AI for long coding tasks and for agents that run many sub-agents. It also reads images and video. The license is the MIT license with one extra sentence that only affects very large products. It needs a server-class GPU node, and Moonshot has since released K2.7 Code and K3.
- Can I use Kimi K2.6 commercially?
- Yes, with conditions. Kimi K2.6 is released under the Modified MIT, which allows commercial use but adds extra rules: Attribution or model name must be shown. Read the license text before you ship a product.
- Is Kimi K2.6 open source?
- Kimi K2.6 is open-weight, not open source in the strict sense. You can download the model files, but the Modified MIT adds conditions that the Open Source Initiative definition does not allow.
- Can I run Kimi K2.6 on my own computer?
- Not on a desktop. The smallest variant, K2.6 (INT4), needs about 655 GB at 4-bit precision, which means a GPU server with several data-center cards.
- When was Kimi K2.6 released?
- Kimi K2.6 was released by Moonshot AI on April 20, 2026.
- How long a context window does Kimi K2.6 have?
- Kimi K2.6 supports a context window of up to 256K tokens.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- Kimi-K2.6 model card (Hugging Face) huggingface.co
- Kimi K2.6 tech blog: Advancing open-source coding (Moonshot AI) kimi.ai
- Agent Swarm capabilities, with the K2.5, K2.6 and K3 release dates (Kimi Help Center) kimi.com
- Kimi K2.5: Visual Agentic Intelligence (technical report the K2.6 card cites) arxiv.org
- Kimi K2.6 license (Modified MIT, Hugging Face) huggingface.co
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Kimi K2.7 Code Moonshot AI | 1.03T 32B active | 256K | Modified MIT (Conditional) | 655 GB GPU server | |
| Kimi K3 Moonshot AI | 2.8T 104B active | 1M | Kimi K3 License (Conditional) | 1.72 TB Multi-node cluster | |
| Kimi K3.1 Pro Moonshot AI | 2.8T 104B active | 1M | Kimi K3 License (Conditional) | 1.72 TB Multi-node cluster | |
| Kimi K2.5 Moonshot AI | 1.03T 32B active | 256K | Modified MIT (Conditional) | 655 GB GPU server |