Moonshot (Kimi)
Moonshot AI develops the Kimi series of models, specializing in long-context understanding. Based in China, Moonshot has gained recognition for its strong performance on long-document tasks. The API uses an OpenAI-compatible format with Bearer authentication.
Getting an API Key
- Visit platform.moonshot.cn/console/api-keys
- Sign in or create a Moonshot account
- Generate a new API key (starts with
sk-...) - Paste the key into AISCouncil under Settings > AI Model > Moonshot
Moonshot does not offer a free API tier. All usage is pay-as-you-go. See the Moonshot platform for current pricing.
API keys are stored locally in your browser (localStorage) and are never included in shared bot URLs.
Supported Models
| Model | Context | Max Output | Input Price | Output Price | Capabilities |
|---|---|---|---|---|---|
| Moonshot v1 128K | 128K | 4K | $0.84/MTok | $0.84/MTok | Streaming |
| Moonshot v1 32K | 32K | 4K | $0.34/MTok | $0.34/MTok | Streaming |
Prices are per million tokens (MTok). Both models have equal input and output pricing.
- Moonshot v1 128K is the primary model -- use it when you need the full 128K context window for long documents.
- Moonshot v1 32K is more affordable and sufficient for shorter conversations and tasks.
Long Context Specialist
Moonshot's Kimi models are specifically optimized for long-context understanding. With a 128K context window and competitive pricing, Moonshot v1 128K is well-suited for:
- Analyzing long documents, reports, or articles
- Summarizing extensive text
- Question-answering over large bodies of text
Configuration
Select Moonshot as the provider when creating a bot profile. The app connects to api.moonshot.cn/v1 using the OpenAI-compatible format with Bearer authentication.
Tips for Best Results
- Use Moonshot for long-document tasks in Chinese and English. The models are optimized for both languages and handle long input efficiently.
- Choose the 32K model when context needs are modest. It costs less than half the price of the 128K model for the same quality on shorter inputs.
- Note that tool calling is not supported. If your workflow requires function/tool calling, consider pairing Moonshot with another provider in a council.