Kimi K3

Kimi K3 is a multimodal model from Moonshot AI (Kimi) with a 1M token context and toggleable max-effort thinking, designed for long-horizon agent work. It is the first version of the kimi-k3 line in the registry. The model has a context window of 1,048,576 tokens and a maximum output of 1,048,576 tokens. Pricing starts at $0.72 per million input tokens and $8.00 per million output tokens, with cached input at $0.195 per million tokens. The weights are open. It accepts image, text, and video input, returns text, and supports reasoning mode and tool calling.

First version of its line in the registry.

Range across the hosts that serve it: up to $6.00 per 1M input tokens.

• ·1M token context window
• ·1M token maximum output
• ·Toggleable max-effort thinking
• ·Accepts image, text, and video
• ·Open weights with tool calling

• Reach for it for long-horizon agent tasks
• Reach for it for processing large multimodal inputs
• Reach for it for extended reasoning with tool use
• Reach for it for tasks needing 1M token output

68 hosts serve this model at the price above · the maker's documentation

• 8Ember-12026-09-28
• 9Google TPU Runs Kimi 57% Faster Than Nvidia GPU, Using DeepSeek Inference Framework2026-09-27
• 8OpenRouter Analyzes Kimi K3: Open Weights, License Terms, and How to Call It2026-09-24
• 8Artificial Analysis 评测 Step 5 Preview： Intelligence Index 得 44 分，成本约为同级模型 1/2.82026-09-22

4 signals of the archive mention this model, each scored against the published bar. This list is rebuilt from the archive at every build, so it cannot go stale.

The registry keeps the six most recent versions of a line. Older ones stay in the file and can be surfaced without a rebuild.

What is the context window size?

The context window is 1,048,576 tokens, and the maximum output is also 1,048,576 tokens.

Is Kimi K3 open weights?

Yes, the weights are open. It is served by 68 hosts.

What input types does it support?

It accepts image, text, and video input, and returns text. It also supports reasoning mode and tool calling.