Kimi K3.1 Pro
Moonshot's refined 2.8T-parameter model with open weights, 1M-token context and image input, tuned for coding and agent work.
In short. Kimi K3.1 Pro is an upgraded version of Kimi K3, one of the largest AI models anyone can download. It reads text and images, can keep about a million tokens (several novels) in view, and is built to carry out long coding and office tasks on its own. It is far too big for a home computer, so most people will use it through an API.
Summary
Kimi K3.1 Pro is Moonshot AI's follow-up to Kimi K3. Moonshot calls it its most capable 3T-class open-weight model. It is post-trained from K3 with large-scale reinforcement learning across hundreds of downstream domains, and it is the first model in the K3.1 series. The weights were published on 2 November 2026.
It is a mixture-of-experts model with about 2.8 trillion parameters in total, of which roughly 104 billion are active for each token (16 of 896 routed experts, plus 2 shared experts). Most attention layers use Kimi Delta Attention, a linear-attention design. Compared with K3, Moonshot replaced the full-attention layers with Gated DeepSeek Sparse Attention, which makes very long inputs cheaper to process. The model accepts text and images, writes text, and has a context window of 1,048,576 tokens.
The weights are distributed in MXFP4, a 4-bit format the model was trained with from the start, so the 1.56 TB download is already the compact version. Moonshot reports gains over K3 in coding, 3D creation, financial research and factual accuracy. Those numbers come from Moonshot's own model card and harnesses and have not been reproduced independently. The license is the Kimi K3.1 Pro License, which carries the same terms as the Kimi K3 License.
Good for
- Self-hosting a frontier-class open model for long coding and agent workloads
- Reading documents, charts and screenshots together with tool use
- Organizations that need full control over their weights and data and already run multi-GPU servers
- Research on very large sparse mixture-of-experts models
Watch out for
- The weights alone are about 1.56 TB, so it needs a multi-GPU data-center setup and is not a local model
- Hosting businesses with over US$20 million combined revenue need a separate agreement, and very large products must credit the model
- Benchmark results are Moonshot's own, measured with the harnesses it chose
- It reads images but cannot hear audio or generate images
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Estimates for the K3.1 Pro 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 |
|---|---|---|---|---|---|
| K3.1 Pro MXFP4 | 2.8T 104B active | 6.44 TB | 3.22 TB | 1.72 TB | Multi-node cluster |
Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.
The license in plain words
An MIT-style license with two scale conditions. Most companies can use it commercially without asking, but large hosting businesses need a deal and very large products must credit the model. Kimi K3.1 Pro uses the same terms under its own name.
- Separate deal needed to sell hosted access above a threshold
- Attribution or model name must be shown
See also
- Kimi K3.1 Pro vs Qwen3.8-Max: which giant open model fits you?
- Best open-source LLMs in 2026
- Best open-source models for coding and AI agents
- Meituan's LongCat-2.5: 1.6T parameters, vision and the plain MIT license
- Giant open models now ship in 4-bit: what it means for memory planning
- Open licenses: free to build on, but hosting needs a deal
- Moonshot releases Kimi K3.1 Pro: 2.8T parameters, open weights
Frequently asked questions
- What is Kimi K3.1 Pro?
- Kimi K3.1 Pro is an upgraded version of Kimi K3, one of the largest AI models anyone can download. It reads text and images, can keep about a million tokens (several novels) in view, and is built to carry out long coding and office tasks on its own. It is far too big for a home computer, so most people will use it through an API.
- Can I use Kimi K3.1 Pro commercially?
- Yes, with conditions. Kimi K3.1 Pro is released under the Kimi K3 License, which allows commercial use but adds extra rules: Separate deal needed to sell hosted access above a threshold; Attribution or model name must be shown. Read the license text before you ship a product.
- Is Kimi K3.1 Pro open source?
- Kimi K3.1 Pro is open-weight, not open source in the strict sense. You can download the model files, but the Kimi K3 License adds conditions that the Open Source Initiative definition does not allow.
- Can I run Kimi K3.1 Pro on my own computer?
- Not realistically. The smallest variant, K3.1 Pro, needs about 1.72 TB at 4-bit precision, more than a single GPU server holds. It is meant for clusters or hosted services.
- When was Kimi K3.1 Pro released?
- Kimi K3.1 Pro was released by Moonshot AI on November 2, 2026.
- How long a context window does Kimi K3.1 Pro have?
- Kimi K3.1 Pro supports a context window of up to 1M tokens.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- Kimi-K3.1-Pro model card (Hugging Face) huggingface.co
- Kimi K3.1 Pro quickstart (Kimi API platform) platform.kimi.ai
- Kimi K3 License huggingface.co
Release date and facts last verified November 8, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Kimi K3 Moonshot AI | 2.8T 104B active | 1M | Kimi K3 License (Conditional) | 1.72 TB Multi-node cluster | |
| Qwen3.8-Max Alibaba Qwen | 2.45T 95B active | 1M | Qwen3.8-Max License (Conditional) | 1.69 TB Multi-node cluster | |
| GLM-5.3 Z.ai (Zhipu AI) | 753B 40B active | 1M | GLM-5.3 License (Conditional) | 520 GB GPU server | |
| Step 5 Preview StepFun | 604B 27B active | 1M | StepFun 1.0 (Conditional) | 417 GB Big-memory desktop |