Kimi K3
Moonshot AI · #2 in our ranking · Most capable
Kimi K3 is an open-weight mixture-of-experts model from Moonshot AI, released Jun 2026, with 2.8T parameters, a 1M-token context window, and the Kimi K3 License.
Kimi K3 is a 2.8T-parameter mixture-of-experts model that activates 16 of 896 experts per token. Moonshot built it for long-horizon coding and agentic knowledge work, and it reads text, images, and video in the same model.
Kimi K3 specs
- Provider
- Moonshot AI
- Architecture
- Mixture-of-experts
- Parameters
- 2.8T total
- Context
- 1M tokens | 786K words
- License
- Kimi K3 License
- Input
- Text, Image, Video
- Released
- Jun 2026
- Weights
- Hugging Face
Sourced from the model card on Hugging Face. Last checked 2026-10-01.
Highlights
- 2.8T parameters
- Text, image, and video input
- 1M-token context
Where it fits for payment teams
- Partner-bank RFIs
#1 pick. Native text, image, and video input with a 1M-token context, so it reads the whole RFI packet in one pass.
- Chargebacks and disputes
#2 pick. The most capable multimodal open model for complex, document-heavy cases.
- Merchant underwriting
#2 pick. Reads website screenshots and documents natively, useful for spotting fake storefronts.
Harnesses in Runtime
Run Kimi K3 through any of these harnesses on Runtime.
Provider-reported benchmarks
| Terminal Bench 2.1 | 88.3 |
|---|---|
| Toolathlon Verified | 76.5 |
Benchmark scores are reported by each model provider on its Hugging Face model card. They are not independently verified by Runtime.
How to run Kimi K3
Download the weights
Pull moonshotai/Kimi-K3 from Hugging Face. Check the Kimi K3 License before you deploy.
Serve it in your cloud
Run it with vLLM or SGLang on your own GPUs, or use a managed provider that hosts the model.
Put it to work in Runtime
Point an agent or a single skill at the model through OpenCode, then compare it to your current model with evals.
Kimi K3 FAQ
What is the context window of Kimi K3?
+
Kimi K3 supports a 1M-token context window (1,048,576 tokens), roughly 786K words of English text.
How many parameters does Kimi K3 have?
+
Kimi K3 is a mixture-of-experts model with 2.8T total parameters.
Is Kimi K3 free for commercial use?
+
Kimi K3 is released under the Kimi K3 License, a custom license. Commercial use may come with conditions, so read the license on Hugging Face before you deploy or fine-tune it.
What can Kimi K3 take as input?
+
Kimi K3 accepts text, image, video input and generates text.
Is Kimi K3 good for partner-bank RFIs?
+
Yes. Kimi K3 is our #1 pick for partner-bank RFIs. Native text, image, and video input with a 1M-token context, so it reads the whole RFI packet in one pass.
Is Kimi K3 good for chargebacks and disputes?
+
Yes. Kimi K3 is our #2 pick for chargebacks and disputes. The most capable multimodal open model for complex, document-heavy cases.
Is Kimi K3 good for merchant underwriting?
+
Yes. Kimi K3 is our #2 pick for merchant underwriting. Reads website screenshots and documents natively, useful for spotting fake storefronts.
Can I run Kimi K3 in my own cloud with Runtime?
+
Yes. Download the weights from Hugging Face (moonshotai/Kimi-K3) and serve them with vLLM or SGLang in your cloud, or use a managed provider that hosts the model. Runtime agents can then use Kimi K3 through a harness like OpenCode, with your data staying in your environment.