Back to models
Chat Models

Kimi

Kimi K2.6

Best Overall

Multimodal Kimi with 256K context and optional thinking for coding, document analysis and visual review.

256K contextOptional thinking
Open AppBrowse All Models

At a glance

Know the model before you prompt.

Kimi API specifications
Context window
262,144 tokens
Maximum output
Not verified
Inputs → output
Text + Images + Video → Text
Knowledge cutoff
Not verified

API model ID: kimi-k2.6

Capabilities & boundaries

What it supports. Where the limits are.

Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.

Supported API features and tools

  • A 262,144-token context for supplied code and documents
  • Thinking enabled by default with an optional non-thinking mode
  • Native text, image and video understanding with text output
  • Tool workflows and multi-turn assistance through a configured integration

Before you choose

  • The model quickstart's 32,768 max_tokens value is a default, not a verified maximum output. A model-specific output ceiling and knowledge cutoff were not established; neither is borrowed from K3.
  • Vision and video understanding produce text; they do not imply native image, video or audio generation. Check what the app accepts before planning a multimodal workflow.
  • Public image URLs are not accepted by the documented API workflow. Use supported encoded content or uploaded file references, and follow the model-specific image and video guidance.
  • Tools require an integration to execute them. Validate arguments and results, and require approval before writes, purchases or other consequential actions. No model benchmark guarantees success on your workflow.

Extended thinking

enabled · defaultdisabled

K2.6 supports thinking.type enabled or disabled, with thinking enabled by default. Its documented temperature is fixed at 1.0 while thinking and 0.6 without thinking; top_p is 0.95. Test both modes on the same acceptance criteria rather than assuming extra thinking is always necessary.

  • The model quickstart lists 32,768 as the max_tokens default, not a maximum. The current API reference favors max_completion_tokens; follow its model-specific request schema instead of treating an older example as the output ceiling.
  • In thinking workflows, tool_choice supports auto or none. Preserve returned reasoning_content with assistant history as the provider recommends; omission can reduce performance. Do not assume K3's required tool choice, dynamic loading or strict schema features apply.
  • K2.7 Code fixes temperature at 1.0; K2.6 uses its mode-specific values. Both document top_p 0.95, n 1 and zero penalties. Follow the specific model rather than copying a generic sampling configuration.
  • Older quickstarts warn about the built-in web-search workflow. The later platform changelog introduces separate Search and Search Pro APIs; these are distinct services, not automatic search access inside a chat prompt or an EZ Ai Assist feature guarantee.

Put it to work

Start with a more useful prompt.

Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.

Workflow 01

Select a thinking mode with evidence

Find the least costly mode that meets the requirement.

Group these real support and coding tasks by difficulty and design a test of thinking enabled versus disabled. Define pass criteria, ambiguity handling and a table for latency and usage. Explain which failures require escalation to a person. Do not assume the thinking mode wins or invent benchmark results.

Workflow 02

Check a long-running implementation plan

Keep progress tied to the original request.

Review this implementation plan, progress log and supplied test results. Identify completed steps supported by evidence, unfinished work and scope drift. Suggest the smallest next checkpoint and the test that would validate it. Keep assumptions visible, preserve unrelated work and do not label the feature complete until all acceptance criteria have evidence.

Workflow 03

Summarize a visual incident timeline

Combine observations without inventing causality.

Using these screenshots, timestamps and incident notes, build a timeline of observed events. Cite the supporting item for each event, distinguish sequence from cause, and mark contradictions or missing intervals. Propose two checks to distinguish the leading explanations. Do not infer private system state that the supplied evidence does not show.

Developer reference

Kimi API pricing

These are Kimi API reference prices, not EZ Ai Assist subscription prices.

View EZ Ai Assist plans
Standard processing · USD per 1,000,000 tokens
Token typePrice
Input · cache miss$0.95
Input · cache hit$0.16
Output$4.00
  • Prices are per 1,000,000 tokens and exclude applicable taxes. Input and output are billed separately; inspect actual usage rather than estimating tokens from character counts.
  • Only eligible cached input receives the cache-hit rate. Uncached input and output keep their own rates. K3's separately published cache-write prices must not be assumed for K2 models.
  • These are direct Kimi platform rates, not universal reseller prices. Recheck the provider's current table and account limits before budgeting a deployment.

Common questions

A few things worth knowing.

What is the default thinking mode?

Thinking is enabled by default and can be disabled using the documented thinking object. The quickstart specifies different fixed temperatures for the two modes; do not treat sampling controls as unrestricted.

Is K2.6 interchangeable with K2.7 Code?

No. They have separate IDs, documented behavior and cached-input rates. K2.6 allows non-thinking requests; K2.7 Code does not. Evaluate each against the tasks you actually run.

Can a prompt enable web search?

No. Search requires a supported integration. Older quickstarts describe restrictions on built-in search; newer standalone Search APIs are separate services. Neither means search is automatically available in your app workspace.

Why is maximum output marked unverified?

The quickstart's 32,768 figure is a default completion setting, while the model-specific maximum is not established in the reviewed sources. Context capacity and maximum output are different limits.

Does the context window guarantee complete recall?

No. Context is a capacity limit, not a retrieval-accuracy guarantee. Organize documents, label evidence and ask for references to the supplied material. Evaluate omissions and contradictions before trusting a long answer.

Can it create images or videos?

The documented model accepts text, image and video inputs and returns text. That supports visual analysis, not native image or video generation. Input support in your app workspace must be checked separately.

Are these API capabilities included in my EZ Ai Assist plan?

This guide describes the provider API, not subscription entitlements or a promise that every setting is exposed in the app. Check your workspace for model access and supported inputs, and use the EZ Ai Assist pricing page for plan details.

Check the source

Official documentation

Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.

Same provider

More from Kimi

Kimi

Kimi K3

Multimodal Kimi with 1M context, always-on thinking and adjustable reasoning effort for coding and knowledge work.

Kimi

Kimi K2.7 Code

Coding-focused Kimi with 256K context, visual inputs and always-on thinking for multi-step software work.

Kimi

Kimi K2.7 Code HighSpeed

Faster-serving K2.7 Code endpoint with the same model behavior, 256K context and separately priced API usage.

Kimi

Kimi K2.5

Historical multimodal Kimi with 256K context, retired on the Kimi platform August 31, 2026. Migration reference for K3.