Kimi is the AI assistant from Moonshot AI. Its open-source K2 model drew a lot of attention, and Kimi K3 is now the flagship with a 1M-token context window.
1. What is Kimi?
Kimi is an AI assistant built by Moonshot AI, a company based in China. You can use it through the Kimi app and website, and developers can call its models through an API. It competes with general assistants such as Grok, ChatGPT, Gemini and Claude.
Moonshot's developer platform now lives at platform.kimi.ai, which redirects from the older platform.moonshot.ai address, seen July 2026. If you have old links or guides pointing at moonshot.ai, expect them to move.
2. Why did Kimi K2 go viral?
The attention came mostly from Kimi K2, which Moonshot released as an open-weight model and described as a powerful open-source agent. Open weights mean the model can be downloaded and run on your own hardware or a third-party host, rather than only through one company's app.
That matters because it lets developers, researchers and cost-sensitive teams test a strong model without being locked to a single provider. Grok, by contrast, is used through xAI's own product and API, not as a downloadable model.
The reason Kimi spread so fast is simple: a capable model that teams could download and run themselves, not just rent through an app.
3. The current Kimi models
As of July 2026, the official platform lists a small set of models. The exact prices should be checked on the official pricing pages before you rely on them.
Kimi K3 (flagship)
Kimi K3 is the current flagship, with a 1M-token context window, confirmed on the official platform. Reported figures elsewhere describe it as a very large model with native vision and pricing in the region of USD 3 input and USD 15 output per million tokens, with a lower cached-input rate. Treat those numbers as reported and verify them at platform.kimi.ai.
Kimi K2.7 Code and K2.6
The platform also lists Kimi K2.7 Code, a dedicated coding model described as multimodal, and Kimi K2.6, which supports visual and text input. The older Moonshot V1 series is being retired, with a full platform sunset noted for the end of August 2026.
4. How does Kimi compare with Grok?
At a high level, Grok leans on live context from X and the web and is delivered as a polished consumer assistant plus an xAI API. Kimi leans on very long context, coding and the flexibility of open weights for the K2 line.
Neither is simply better. If live public conversation and a single managed assistant matter most, Grok is the more natural starting point. If you want a long-context or coding-focused model, or the option to self-host, Kimi is worth testing. For a side-by-side view, see the full Grok vs Kimi comparison.
5. How to try Kimi
For everyday use, open kimi.com or the Kimi app and give it one real task, the same way you would test any assistant. For software, create an API key at platform.kimi.ai, which uses pay-as-you-go billing with no subscription plan, seen July 2026.
Whichever route you take, verify current model names, limits and prices on the official pages, because AI product pages change often and third-party guides go stale quickly.