Kimi K3
At 2.8T total parameters with 104B active, Kimi K3 is the largest open-weight model in general circulation and the only one here that ingests video natively. Reasoning is always on with configurable effort. The licence is the catch: it is not a standard open licence, and organisations above $20M in revenue need a separate commercial agreement.
Pricing checked against provider documentation on . How we verify
kimi-k3 reasoning open weights Input
$3.00
/ 1M tokens
Output
$15.00
/ 1M tokens
Context
1.05M
1,048,576 tokens
Max output
131K
tokens per request
Pricing caveat: The custom licence requires a separate commercial agreement once the licensee and its affiliates exceed $20M revenue over any consecutive 12 months.
What it costs in practice
A workload of 10M input and 2M output tokens per month — roughly a moderately busy production assistant — runs $60.00/month on Kimi K3. The next cheaper option, GPT-5.6 Terra , would cost $44.00.
Where Kimi K3 fits
- Multimodal agent workflows
- Video and image understanding at length
- Ambitious long-horizon automation
- Research requiring open weights at scale
Specifications
- API model ID
- kimi-k3
- Provider
- Moonshot AI
- Model family
- Kimi K
- Released
- June 2026
- Knowledge cutoff
- February 2026
- Context window
- 1,048,576 tokens
- Max output
- 131,072 tokens
- Native reasoning
- Yes
- Relative latency
- high
- Input modalities
- text, image, video
- Output modalities
- text
- Open weights
- Kimi K3 License (commercial terms above $20M revenue)
- Cached input
- $0.30 / 1M tokens
- Batch discount
- —
Frequently asked
How much does Kimi K3 cost?
Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens, with cached input at $0.30.
What is Kimi K3's context window?
Kimi K3 accepts up to 1,048,576 tokens of context and can generate up to 131,072 output tokens per request.
Is Kimi K3 a reasoning model?
Yes. Kimi K3 performs native chain-of-thought before answering. Those thinking tokens are billed at the output rate, so budget above the sticker price.
What is Kimi K3 best for?
Natively multimodal long-horizon agents. At 2.8T total parameters with 104B active, Kimi K3 is the largest open-weight model in general circulation and the only one here that ingests video natively. Reasoning is always on with configurable effort. The licence is the catch: it is not a standard open licence, and organisations above $20M in revenue need a separate commercial agreement.