How to Choose: K2.6 / K3 / K3 Cluster, Reasoning Effort, and Agent
Kimi will decide on its own whether your question needs web access, so you do not need to switch it manually. What you do need to choose is the model and reasoning effort; for complex tasks, you may also use Agent and Agent cluster.
Three Model Options
Use the model switcher above the input box to choose:
| Model | Reasoning effort | Best for | Credits |
|---|---|---|---|
| K2.6 | Standard / Advanced | Fast dialogue and Q&A, with faster responses | Does not consume member credits in chats |
| K3 | Standard / Advanced / Extreme | Conversations and Agent tasks, with the strongest overall capability | Billed against credits |
| K3 Cluster | Standard / Advanced / Extreme | Large-scale search, batch processing, and completing high-volume tasks in one go | Billed against credits |
- K3 can produce end-to-end editable documents, such as
.pptx,.docx,.xlsx, and.pdf; K2.6 is better suited for fast Q&A and only outputs text content, such as PPT outlines, Word body text, and code. - To generate images / videos / audio, click “+” → Plugins, select the corresponding feature, and use it with K3.
K2.6 is free and does not consume credits only when used from the chat entry point (the model selector above the input box). When used as a K2.6 Agent in Kimi Work, it still consumes member credits.
Reasoning Effort: Standard / Advanced / Extreme
The higher the reasoning effort, the more thoroughly the model reasons and, in general, the smarter it becomes—but it also uses more tokens.
- If you want speed and lower credit usage → choose Standard;
- If you need stronger reasoning for complex analysis or difficult problems → choose Advanced or Extreme (Extreme is supported by K3 / K3 Cluster).
Agent and Agent Cluster
- General Agent: Automatically plans and completes tasks, including website generation, PPT creation, Deep Research, and document and spreadsheet processing.
- Agent cluster: Supports more than 4,000 parallel tool calls and can autonomously coordinate up to 300 subagents (Sub-agents) for parallel processing. It is suitable for large-scale tasks such as high-volume search, long-form writing, and batch processing.
Which One Should I Choose?
- fast Q&A and everyday chat → K2.6 (does not consume credits in chats).
- Complex conversations, writing documents / creating PPTs and spreadsheets / multi-step tasks → K3.
- Large-scale search, batch processing, and ultra-long-form writing → K3 Cluster.
- To save credits → use K2.6, or set reasoning effort to “Standard”.
FAQ
Is “Thinking mode” no longer available?
Kimi has merged “chat” and “Agent” into one experience and no longer separates them into different modes. “Thinking” is now reflected through your choice of model and reasoning effort—for example, using K3 with a higher reasoning effort gives you more thorough reasoning.
What is the difference between K2.6 and K3?
K2.6 is better suited for fast Q&A, responds faster, and does not consume member credits in chats. K3 has the strongest overall capability, excels at complex conversations and Agent tasks, can produce end-to-end editable documents, and is billed against credits.
What is the difference between K3 and K3 Cluster?
K3 is suitable for general conversations and Agent tasks. K3 Cluster uses large-scale parallelism—thousands of tool calls and hundreds of subagents—to handle high-volume tasks such as large-scale search, batch processing, and ultra-long-form content.
How much content can be processed in a single turn?
K2.6 supports about 128K tokens in a single-turn context (roughly 50,000–60,000 Chinese characters). K3 provides a 1-million-token context window (requires the highest-tier membership benefit). See What should I do if my chat exceeds 200,000 Chinese characters?.