Moonshot AI · Open weights (custom)
Kimi K3 API
The largest open-weights model — native vision and multi-agent Swarm.
Context: 1M tokens · Released: July 2026
# Kimi K3 via kryyt (OpenAI SDK) from openai import OpenAI client = OpenAI(base_url="https://api.kryyt.com/v1", api_key="$KRYYT_API_KEY") r = client.chat.completions.create( model="kimi-k3", messages=[{"role":"user","content":"Explain MoE routing in one paragraph."}], ) print(r.choices[0].message.content)
Overview
Kimi K3 from Moonshot AI is the largest open-weights model released to date, a mixture-of-experts model with native vision, a 1M-token context, and a Swarm system that orchestrates hundreds of specialized sub-agents across long autonomous runs.
Best for:
- Long-horizon autonomous agents
- Multimodal image + video input
- Research and multi-step orchestration
Kimi K3 benchmark performance
Kimi K3 leads the open models here on GPQA Diamond and BrowseComp, and is near the top on agentic coding. Moonshot's coding results use its own Kimi Code harness.
| Benchmark | Model | Score (%) |
|---|---|---|
| GPQA DiamondGraduate-level science reasoning | Kimi K3 | |
| BrowseCompAutonomous web research | Kimi K3 | |
| Terminal-Bench 2.1Agentic coding in real terminal environments | Kimi K3 | |
| MMMU-ProMultimodal, college-level reasoning | Kimi K3 | |
| FrontierSWEFrontier-difficulty software engineering | Kimi K3 | |
| DeepSWE v1.1Repairing fresh, real-world software issues | Kimi K3 | |
| Humanity's Last ExamFrontier knowledge and reasoning (no tools) | Kimi K3 |
Sources: Moonshot AI Kimi K3 model card (opens in new tab).
Scores are vendor-reported unless marked independent, as of September 2026. Labs use different agent harnesses, reasoning settings, and benchmark versions, so treat cross-lab gaps of a few points as roughly equal and test on your own workload. See all models on the benchmarks page.
Kimi K3 models & pricing
| Model | API ID | Context | Pricing | License | Status |
|---|---|---|---|---|---|
| Kimi K3 | kimi-k3 | 1M | See account pricing | Open weights (custom) | Soon |
kimi-k3
- Model
- Kimi K3
- Context
- 1M
- Pricing
- See account pricing
- License
- Open weights (custom)
- Status
- Soon
Usage is charged at Kryyt's published rates for the selected model and features. Review current pricing and usage in your account.
Kimi K3 FAQ
What is Kimi K3?
Kimi K3 from Moonshot AI is the largest open-weights model released to date, a mixture-of-experts model with native vision, a 1M-token context, and a Swarm system that orchestrates hundreds of specialized sub-agents across long autonomous runs.
How do I call Kimi K3 on kryyt?
Send a POST request to https://api.kryyt.com/v1/chat/completions with your kryyt key and set "model" to one of: kimi-k3.
What license is Kimi K3 under?
Kimi K3 is available under the Open weights (custom) license via its provider, Moonshot AI.
What context window does Kimi K3 support?
Kimi K3 supports 1M tokens depending on the specific model variant.
Ship with one endpoint
Get a key, change your base URL, and call every model on this page. No new SDKs.