Kimi K3
Kimi K3 is Moonshot AI's July 2026 multimodal model for long-horizon coding, reasoning, and knowledge work.
Provider
Moonshot AI
Model family
Moonshot Kimi
Multimodal MoE LLM
Cost tier
Flagship
Status
Current
Release Jul 16, 2026
Why teams choose it
Use Kimi K3 as the Kimi-family candidate for long-context coding and research workflows
then evaluate on complete multi-step tasks rather than isolated prompts.
The Kimi API documents low, high, and max reasoning-effort settings; measure quality, la…
tency, and token use for each lane.
Tradeoffs to know
- A one-million-token limit does not guarantee uniform recall across the entire window; run position-aware long-context evals.
- Verify checkpoint availability, license terms, API quotas, pricing, and regional access before production deployment.
When not to use this
- Not ideal for simple tasks where cheaper models in the same lineup are good enough.
- Avoid for latency-sensitive real-time chat when raw response speed outweighs reasoning depth.
- Confirm limits, pricing, and regional availability on the provider side before committing production workloads.
Technical specs
- Inputs
- text, image
- Outputs
- text
- Capabilities
- reasoning, agentic coding, native vision, long context, tool use
- License
- Open weights announced; verify the published Kimi license
- Model string
kimi-k3
Benchmarks
No benchmark data yet.
Kimi K3 FAQ
What is Kimi K3?
Kimi K3 is Moonshot AI's multimodal reasoning and coding model released in July 2026. Kimi documents native vision, a one-million-token context window, and access through Kimi products and API.
Is Kimi K3 open weight?
Moonshot announced full Kimi K3 weights for release, but teams should verify that the published checkpoint and license match their intended commercial, hosting, and redistribution use before deployment.
Explore next
Models, tools, and comparisons that connect to this reference.