KimiK3 MaxKimi K3: The Independent Guide
Kimi K3 is Moonshot AI's 2.8-trillion-parameter flagship — the largest open-weight model announced to date. This guide breaks down what it is, how it works, how it performs, what it costs, and how to use it.
What is Kimi K3?
Kimi K3 is the flagship model from Beijing-based Moonshot AI, the lab behind the Kimi assistant. Announced on July 16, 2026, it is described by Moonshot as the world's first open model in the 3-trillion parameter class. Rather than iterating on the prior K2 design, K3 rebuilds the attention stack and nearly triples total parameters.
Key specifications
| Spec | Kimi K3 |
|---|---|
| Developer | Moonshot AI |
| Released | July 16, 2026 |
| Architecture | Mixture-of-Experts (Stable LatentMoE) |
| Total parameters | ~2.8 trillion (reported) |
| Active experts | 16 of 896 per token |
| Context window | 1,000,000 tokens |
| Input types | Text, image, video |
| API pricing | $0.30 / $3.00 / $15.00 per 1M tokens |
| Weights | Open weights expected July 27, 2026 (Modified MIT) |
Architecture in plain terms
K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes). KDA uses cheaper linear attention for most of the model with periodic full-attention layers, cutting KV-cache memory at long context. The MoE design (Stable LatentMoE) activates only 16 of 896 experts per token, keeping inference practical at 2.8T scale. On the Artificial Analysis Intelligence Index, K3 scores ~57, placing fourth among 189 models.
Explore the clusters
- Specs & Architecture — deep dive into KDA, parameters, and context.
- Benchmarks — coding, agentic, and intelligence-index results vs rivals.
- Pricing & API — token costs and a live calculator.
- How-To Guides — practical walkthroughs.
- Open Weights — weights, licence, and self-hosting.
- FAQ — answers to the most-asked Kimi K3 questions.
- Playground — try Kimi K3 in your browser right now.
Frequently Asked Questions
What is Kimi K3?
Kimi K3 is Moonshot AI's flagship large language model, released July 16, 2026. It uses a Mixture-of-Experts design with a reported 2.8 trillion total parameters and a 1,000,000-token context window, and Moonshot plans to release it as an open-weight model.
Is Kimi K3 open source?
Moonshot describes K3 as an open-weight model. The full model weights were scheduled for release on July 27, 2026, under a Modified MIT licence, but the final licence terms should be verified on the official release.
How much does Kimi K3 cost?
Official API pricing is $0.30 per million cache-hit input tokens, $3.00 per million cache-miss input tokens, and $15.00 per million output tokens. See our pricing page and token calculator for estimates.
Is KimiK3 Max affiliated with Moonshot AI?
No. KimiK3 Max is an independent fan and reference site. It is not affiliated with, endorsed by, or operated by Moonshot AI.