Long-context Kimi K3 model
Kimi K3 provides a 1M-token context window and native visual understanding for large codebases, extensive documents, and complex reasoning workflows.
Kimi API Platform
Freemium
Kimi API Platform gives developers API access to Kimi language models for software engineering, dialogue, visual understanding, reasoning, and tool-using applications. It includes models with context windows up to 1M tokens and usage-based token billing.
What is Kimi API Platform?
Kimi API Platform is a developer API platform for integrating Kimi language models into AI applications. The available models support software engineering, long-context reasoning, dialogue, visual understanding, and agent-oriented workflows.
Kimi K3 is the flagship model, with 2.8 trillion parameters, native visual understanding, and a 1M-token context window. The platform also provides Kimi K2.7 Code and its High-Speed variant for coding, plus Kimi K2.6 for visual and text input, dialogue, and agent tasks. Developers can select a model based on the context length, modality, and workflow required.
What can Kimi API Platform do?
Coding-focused models
Kimi K2.7 Code is designed for coding tasks and long-context instruction following. The K2.7 Code High-Speed variant is listed with output speed of approximately 180 tokens per second, reaching up to 260 tokens per second in short-context scenarios.
Visual and text input
Kimi K2.6 supports both visual and text input, along with thinking and non-thinking modes for dialogue and agent tasks.
Official tool ecosystem
The platform lists tools for web search, persistent memory, Excel and CSV analysis, Python execution, JavaScript execution through QuickJS, URL extraction, date and time processing, unit conversion, and Base64 encoding and decoding.
Usage-based token billing
Inference is billed by tokens, with separate input and output charges. K3 also has separate cache-write pricing by TTL and a cached-input price for repeated request prefixes.
Use Cases
“Software engineering agents”
Use Kimi K3 or Kimi K2.7 Code for code generation, debugging, refactoring, automated testing, and multi-step development workflows across large codebases.
“Research and analytical assistants”
Combine long-context reasoning with tools such as web search and multi-step tool calling for strategic research, competitive or pricing analysis, and financial research workflows.
“Legal and intellectual-property review”
Process contracts, patents, and other high-stakes documents where terminology, detailed reading, structured analysis, and drafting are important.
“Conversation-data analysis”
Analyze unstructured conversation data for tasks such as customer-intent detection, key-information extraction, public-opinion monitoring, and quality assessment.
“Scientific and technical document workflows”
Use visual and text understanding to extract information from charts, formulas, research literature, and other technical materials.
Frequently Asked Questions
Which Kimi models are currently listed?
The model list includes kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed, and kimi-k2.6. The documentation also identifies several older models as deprecated.
What are the context windows?
Kimi K3 has a 1,048,576-token context window. Kimi K2.7 Code, Kimi K2.7 Code High-Speed, and Kimi K2.6 are listed with 262,144-token context windows where specified by the pricing or model documentation.
How is model inference billed?
Input and output tokens are billed separately. K3 additionally charges for cache writes according to a 5-minute or 1-hour TTL, while repeated prefixes that hit the cache use the cached-input price without another cache-write charge. Prices are stated per 1 million tokens and exclude applicable taxes.
Does the platform support tools?
Yes. The platform lists official tools including web search, memory, Excel analysis, Python and JavaScript execution, URL fetching and extraction, date processing, unit conversion, and Base64 encoding and decoding.
What should developers use the models for?
Kimi K3 is positioned for frontier intelligence tasks such as software engineering, knowledge work, and deep reasoning. Kimi K2.7 Code is dedicated to coding, while Kimi K2.6 supports visual and text input, dialogue, and agent tasks.
Quick Facts
- Product type
- Developer API platform for Kimi language models
- Flagship model
- Kimi K3
- Largest context window
- 1,048,576 tokens
- Available model focus
- Long-context reasoning, coding, visual and text input, dialogue, and agent tasks
- Billing basis
- Usage-based input and output token billing
- Source domain
- platform.kimi.ai
Kimi API Platform Alternatives
OpenRouter
openrouter.ai
OpenRouter gives teams one API for hundreds of AI models, with routing, fallback, OpenAI-compatible access, SDKs, prompt caching, and usage controls.
Sapling
sapling.ai
Sapling is a language-model toolkit and assistant for developers and businesses, with API/SDK access and integrations for writing help, text analysis, and enterprise workflows.
Experiential Labs
www.experientiallabs.ai
Experiential Labs is an open-source AI gateway that gives teams one endpoint and API key for hosted models, customer-owned keys, and local or custom models. It adds routing, budgets, access control, attribution, and optional intelligence features for agent traffic.
EasyRouter
easyrouter.io
EasyRouter is an AI model gateway that unifies access to third-party model APIs through one interface, with standard formats and usage-based billing.
Twigg
twigg.ai
Twigg is a stateful API for building LLM applications, with hosted conversation storage, context assembly, compaction, and model routing. Send only the next event, preserve chats across requests, and switch models or providers without moving conversations into vendor-specific threads.
NavtoAI API
sub2api.navtoai.com
NavtoAI API unifies access to 200+ AI models through one account and API shape, with routing, failover, usage tracking, and centralized team management.