Moonshot Kimi K3 MoE LLM: AI Text Generator
Moonshot Kimi K3 is a flagship 2.8T Mixture-of-Experts (MoE) LLM with a 1M token context window for long-context reasoning, coding, and complex agent workflows.
Moonshot Kimi K3 is a flagship 2.8T Mixture-of-Experts (MoE) LLM with a 1M token context window for long-context reasoning, coding, and complex agent workflows.
About this model
Moonshot Kimi K3 is a flagship 2.8T Mixture-of-Experts (MoE) LLM featuring a massive 1M token context window. Designed for deep reasoning, multi-turn coding, and complex agentic task execution.
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $0.0001 per request | Token-based pricing ($0.70/M input tokens, $2.80/M output tokens). |
Token-based pricing ($0.70/M input tokens, $2.80/M output tokens).
* Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Prompt | string | The user message or instruction for the model. | Explain the architecture of Mixture-of-Experts models. |
| Image URL | string | Optional image URL to include as multimodal input. | undefined |
| System Prompt | string | Optional system-level instruction to guide model behavior. | You are an expert AI researcher and coding assistant. |
| Temperature | number | Controls output randomness (0.0 to 2.0). Lower values are more deterministic. | 1 |
| Max Tokens | integer | Maximum number of tokens to generate. | 4096 |
| Top P | number | Nucleus sampling parameter (0.0 to 1.0). | 1 |
| Frequency Penalty | number | Frequency penalty (-2.0 to 2.0). Reduces likelihood of repeating words. | 0 |
| Presence Penalty | number | Presence penalty (-2.0 to 2.0). Increases likelihood of new topics. | 0 |
The user message or instruction for the model.
Explain the architecture of Mixture-of-Experts models.Optional image URL to include as multimodal input.
undefinedOptional system-level instruction to guide model behavior.
You are an expert AI researcher and coding assistant.Controls output randomness (0.0 to 2.0). Lower values are more deterministic.
1Maximum number of tokens to generate.
4096Nucleus sampling parameter (0.0 to 1.0).
1Frequency penalty (-2.0 to 2.0). Reduces likelihood of repeating words.
0Presence penalty (-2.0 to 2.0). Increases likelihood of new topics.
0Developer documentation
/api/v1/kimi-k3 or stream via /api/v1/kimi-k3/stream.system_prompt, temperature, max_tokens, top_p, frequency_penalty, and presence_penalty.image_url for multimodal processing.Frequently asked
Moonshot Kimi K3 supports up to a 1,000,000 (1M) token context window.
Yes, live real-time token streaming is supported via Server-Sent Events (SSE) at `/api/v1/kimi-k3/stream`.