Kimi K3: What it does and what it costs
moonshotai/kimi-k3Kimi K3 is MoonshotAI’s open-weight model for text, image, and video input with reasoning and tool use.
1.05M tokens of context — about 786,432 words, on a scale that ends at 2M.
What it can do
- Reasoning
- Vision
- Tool use
- Open weights
- Takes text
- Takes image
- Takes video
- Gives text
Released 16 Jul 2026 · checked 3 Oct 2026.
About it
Kimi K3 is an open-weight model from MoonshotAI that can take text, images, and video as input and produce text. It supports reasoning and tools, with a context window of 1,048,576 tokens and a maximum output of 943,718 tokens. Its listed rates are $2.7 per million input tokens and $13.5 per million output tokens.
Review
In a line
Kimi K3 is an open-weight multimodal model from MoonshotAI. It accepts text, images, and video, and responds with text.
What it's good at
The model is suited to complex coding, knowledge work, and workflows where an AI agent handles steps over a longer period. It supports reasoning and tool use, and can work with images as well as text and video.
Price in plain terms
The listed rates are $2.7 per million input tokens and $13.5 per million output tokens. A page of about 500 words costs roughly 0.18 cents to read. A chat turn with about 1,000 tokens in and 500 out costs about 0.95 cents; 1,000 such turns come to about $9.45.
Context and limits
Kimi K3 can take in up to 1,048,576 tokens at once. Its maximum response is 943,718 tokens. It supports text, image, and video input, but its output is text.
Ways to use it
Kimi K3 is available as an open-weight model. The listed model identifier is moonshotai/kimi-k3.
Verdict
Consider Kimi K3 if you need a model that can reason, use tools, and take image or video input, especially for extended work. Its input and response limits may suit tasks that involve a lot of material. It may not fit if you need image or video output, since it produces text, or if you want a model without token-based pricing.
Verdict
Kimi K3 is worth considering for people who need reasoning and tool use across coding, knowledge work, or longer-running agent workflows. It accepts images and video as well as text, and its context window and maximum output let it handle substantial material in one exchange. The output is text only, and the listed prices are based on token use, with output tokens costing more than input tokens. Those who need generated images or video should look elsewhere.
- Open weights are available.
- Accepts text, image, and video input.
- Supports reasoning and tool use.
- Context window is 1,048,576 tokens.
- Maximum output is 943,718 tokens.
- Produces text only.
- Output tokens cost $13.5 per million, five times the input rate.
- Prices depend on token use.
- Pick it ifComplex coding and knowledge work; Longer-running workflows that use tools; Tasks involving text, images, or video
- Skip it ifYou need image or video output.; You want to avoid token-based pricing.
Pricing
The listed rates are $2.7 per million input tokens and $13.5 per million output tokens. Rough examples: about 0.18 cents to read a page, 0.95 cents for one typical chat turn, and $9.45 for 1,000 such turns.
Questions
Who makes Kimi K3?
MoonshotAI makes Kimi K3.
Is Kimi K3 open weight?
Yes. Its weights are available as an open-weight model.
What can Kimi K3 take as input?
It accepts text, image, and video input. Its output is text.
How much does Kimi K3 cost?
The listed rates are $2.7 per million input tokens and $13.5 per million output tokens.
How much can it read at once?
Its context window is 1,048,576 tokens.
How long can its answer be?
Its maximum output is 943,718 tokens.
Can Kimi K3 use tools?
Yes. It supports tool use and reasoning.