Model Selection and Performance Comparison

Kimi API offers multiple models for developers, each with different strengths in capability, speed, and pricing.

Available Models

Visit platform.kimi.com/docs/introduction for the full model list and detailed parameters.

When selecting a model, we recommend evaluating the following dimensions:

Context length: Different models support different maximum context windows. For long-document processing, choose a model with a larger context window.
Response speed: Lightweight models respond faster and are better suited to latency-sensitive scenarios.
Generation quality: Advanced models perform better on tasks such as complex reasoning and creative writing.
Pricing: Choose the most cost-effective model based on your budget and call volume.

Vision models support image input and can be used for image descriptions, OCR, chart interpretation, and similar scenarios:

Each image is billed at a fixed 1024 tokens, regardless of image size or resolution.
Common image formats are supported, including JPEG, PNG, WebP, and more.
Images can be passed in via URL or Base64 encoding.

Video multimodality: Direct input and understanding of video files are not supported yet.
PPT Generation API: The PPT generation feature is not yet available through API.
Deep Research API: The Deep Research feature is not yet available through API.

For these capabilities, please follow platform announcements for the latest updates.