Kimi K3.1 Pro API

Build coding and agent workflows with Kimi K3.1 Pro or Highspeed through EvoLink’s unified API: 1M context, visual input, prompt caching and web search.

Moonshot AIText GenerationAvailable

From $2.000 / 1M input tokens

API docs

Long-document analysisTool useStructured outputPrompt cachingChat + Messages

Production routeLive

Live

Context

1.05M context · 1.05M max output

Best For

Coding, document analysis, tool workflows

Input

Text + images

Output

Text · JSON (structured output) · tool calls

Models Pricing Readme API

Choose Kimi K3.1 Pro

Choose a model ID for your application. Both variants support Chat Completions, Responses, and Anthropic Messages.

Kimi K3.1 Pro

General coding, document analysis, and tool workflows.

Selected

Model ID

kimi-k3.1-pro

Best for

General coding, document analysis, and tool workflows.

Input

$2.000 / 1M

136 cr / 1M

Cache read

$0.200 / 1M

13.6 cr / 1M

Cache write (5m)

$2.000 / 1M

136 cr / 1M

Cache write (1h)

$4.000 / 1M

272 cr / 1M

Output

$10.000 / 1M

680 cr / 1M

Kimi K3.1 Pro Highspeed

Compare Highspeed with Pro on your own prompts to choose the response time and token cost that fit your application.

Model ID

kimi-k3.1-pro-highspeed

Best for

Compare Highspeed with Pro on your own prompts to choose the response time and token cost that fit your application.

Input

$4.000 / 1M

272 cr / 1M

Cache read

$0.200 / 1M

13.6 cr / 1M

Cache write (5m)

$4.000 / 1M

272 cr / 1M

Cache write (1h)

$8.000 / 1M

544 cr / 1M

Output

$20.000 / 1M

1360 cr / 1M

All rates are per 1M tokens, shown in USD and credits, and reflect your account's current pricing. Enter uncached input separately from cache reads, 5m writes, and 1h writes. Do not count a cache write again as ordinary input. Cache reads cost the same on both variants.

Kimi K3.1 Pro pricing

Estimate what one Kimi K3.1 Pro request costs before you integrate. The calculator uses your account's current rates, with official pricing as a reference.

Request calculator

Enter the token mix for one request and the number of successful tool calls. Enter uncached input separately from cache reads, 5m writes, and 1h writes. Do not count a cache write again as ordinary input. Cache reads cost the same on both variants.

Input tokensCache read tokensCache write tokens (5m)Cache write tokens (1h)Output tokensWeb search

Estimated request cost

Kimi K3.1 Pro

USD$0.0090

Credits0.612

Input tokens0.272 cr

Cache read tokens0 cr

Cache write tokens (5m)0 cr

Cache write tokens (1h)0 cr

Output tokens0.34 cr

Minimum charge: 0.01 credits per request.

Budget guide

Approximate requests using the current token mix.

Add credits

$10

About 1062 requests

For quick testing

$50

About 5555 requests

For regular development

$100

About 11111 requests

For production evaluation

Server-side tool rates

Only successful server-side calls are billed per call; failed attempts have no tool fee, but tokens still apply.

Web search$0.0050/ call0.34 cr / call

Kimi K3.1 Pro specs and capabilities

Numbers come from the EvoLink route configuration; capabilities are what the API exposes today.

Context window: 1,048,576 tokens
Output budget: 131,072 by default; up to 1,048,576 subject to remaining context
Input: Text + images
Output: Text · JSON (structured output) · tool calls
Reasoning: low / high / max
Tool use: Function calling with multi-step tool sequences
Prompt caching: Cache write + cache read rates
Server-side tools: Web search, billed per successful call
Protocols: Chat Completions · Responses · Anthropic Messages
Model ID: kimi-k3.1-pro

The two model IDs appear in the selector above. Context is shared by input, reasoning and output; the maximum request setting is not a guaranteed generated length.

What is Kimi K3.1 Pro?

Moonshot describes K3.1 Pro as the first K3.1-series release: a 3T-class reasoning model with stronger long-horizon task execution. Evaluate it on completed-task quality, billed tokens and elapsed time; the 1M context and visual input are shared with K3.

SourcesKimi K3.1 Pro Moonshot official pricing Moonshot integration guide Read the launch discussion

Kimi K3.1 Pro vs K3: What Changed?

Moonshot highlights improvements in coding, 3D creation and financial research. Independent evaluation results follow below.

Kimi K3.1 Pro Coding and Repository Tasks

From repository fixes to complete project builds, the release targets longer tool sequences and iterative debugging. Test a real issue with fixed acceptance tests; compare pass rate, retries and total time against K3.

Kimi K3.1 Pro for 3D Creation

The official examples cover interactive web experiences, game engines, Blender and CAD scripting. The API returns text and tool calls; producing a finished 3D asset or video requires the corresponding tools.

Kimi K3.1 Pro Financial Research and Modeling

Moonshot emphasizes retrieval, cross-document verification and financial modeling. Evaluate source traceability, numerical consistency and linked workbook formulas on a task with a reference answer.

Kimi K3.1 Pro Long-Context Document Tasks

Use repositories, document collections and conversation history to test retrieval, citation accuracy and output completeness. Size the task within the shared context budget rather than assuming a large window guarantees recall.

Kimi K3.1 Pro vs K3: Artificial Analysis Benchmarks

Artificial Analysis launch evaluation · October 23, 2026 · standard Pro, compared with Kimi K3 (max). Scores describe that evaluation configuration.

Metric / unit	Kimi K3.1 Pro	Kimi K3 (max)	Difference / context
Intelligence Index / points	55	44	+11 points
Terminal-Bench 4.0 / success rate	67%	13%	+54 percentage points
GDPval-AA / Elo	1,846 Elo	1,534 Elo	+312 Elo
AutomationBench-AA / success rate	73%	58%	+15 percentage points
Output tokens / Index task	≈ 55,000	≈ 48,000	About 7,000 more tokens
USD / Index task	≈ $1.98	≈ Pro	Roughly equal to K3

First-party input/output token prices fell by about 33%, but more output tokens left measured cost per Index task roughly equal to K3. Data comes from Artificial Analysis’s launch evaluation of standard Pro.

Artificial Analysis · 2026-10-23

Kimi K3.1 Pro vs Highspeed: Speed and Cost

Moonshot describes Highspeed as the same intelligence with about 2–3× faster output. A multi-step agent also waits for tools, builds, search and infrastructure, so this is not a 2–3× task-completion guarantee. Compare cache use, total billed tokens and elapsed time.

Kimi K3.1 Pro API Cost Example (Moonshot Pricing)

Moonshot official USD reference, excluding tax: 100,000 uncached input tokens × $2/1M + 10,000 billed output tokens × $10/1M = $0.30 on Pro, excluding additional services. Reasoning is included in billed output. EvoLink’s calculator uses your current account SKU rates.

Two ways to use Kimi K3.1 Pro: EvoLink API or Agent

Use the EvoLink API for product backends and batch jobs, or call Kimi K3.1 Pro from Codex, Claude, or Gemini for coding and analysis workflows. Both paths share the same EvoLink API key, balance, model ID, and request history.

Option 1

Integrate with the EvoLink API

Best for: product backends, batch jobs, automated pipelines

Send OpenAI-compatible Chat Completions (or Anthropic Messages) requests to EvoLink and control the model ID, system prompt, output budget, tools, and structured output.

1Create an EvoLink API key in the console
2Point your OpenAI or Anthropic SDK at the EvoLink base URL and select the model ID shown above
3Send one representative request and read the usage field for input, cached, and output tokens
4Set max_tokens and retries per task; keep tool-call IDs and results across turns

Create EvoLink API Key

Option 2

Call it with an Agent

Best for: coding, review, and analysis tasks in Codex, Claude, and Gemini

Give the Agent the task, the inputs to include, and the acceptance criteria. It assembles the request, calls Kimi K3.1 Pro through EvoLink, and returns the answer with token usage.

1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
2Describe the task, the inputs to include, and the expected output format
3Ask the Agent to call Kimi K3.1 Pro through EvoLink and show the request before sending
4Let the Agent report the answer, token usage, and any error body

Kimi K3.1 Pro API code example and error handling

This example shows the shortest runnable request: an OpenAI-compatible Chat Completions call with a system prompt, a user message, and an output budget. Open the API tab for the complete parameter and response reference.

View complete API docs

cURL

curl -X POST https://api.evolink.ai/v1/chat/completions \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3.1-pro",
    "messages": [{"role": "user", "content": "Explain this function: <code>"}],
    "max_completion_tokens": 2048,
    "reasoning_effort": "high",
    "prompt_cache_options": {"mode": "implicit", "ttl": "5m"},
    "stream": false
  }'

Invalid request or unsupported parameter

Check the model ID, the messages array, and parameter ranges against the API reference; remove fields this route does not support.

Authentication or balance issue

Check the Authorization bearer token and confirm the available balance in the console.

Context length exceeded

Prompt tokens exceed the Kimi K3.1 Pro context window. Trim or retrieve only the relevant evidence and reuse cached prefixes.

Rate limited (429)

Back off and retry with jitter; batch or queue requests instead of sending parallel bursts.

Content or tool call rejected

Review sensitive content, malformed tool-call arguments, and JSON schema mismatches before retrying.

Kimi K3.1 Pro Python Example: Moonshot Responses API

This example calls Moonshot directly with MOONSHOT_API_KEY, separately from the EvoLink example above. Follow the official API reference. Use kimi-k3.1-pro-highspeed to select Highspeed.

Moonshot direct endpoints: /v1/responses and /v1/chat/completions at https://api.moonshot.ai; Messages at https://api.moonshot.ai/anthropic/v1/messages. EvoLink uses its own base URL and key.

pip install --upgrade openai
export MOONSHOT_API_KEY="YOUR_MOONSHOT_API_KEY"

python - <<'PYTHON'
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MOONSHOT_API_KEY"],
    base_url="https://api.moonshot.ai/v1",
)
response = client.responses.create(
    model="kimi-k3.1-pro",
    input="Review this API integration plan for missing retries and timeouts.",
)
print(response.output_text)
PYTHON

Moonshot Responses API reference

Kimi K3.1 Pro API Parameters and Integration Limits

Fixed Sampling Parameters

Moonshot fixes temperature=1.0 and top_p=0.95, as well as n=1 and zero presence/frequency penalties. Omit these sampling fields instead of carrying settings from another model.

temperature

Assistant Messages and Tool-Call State

For Chat tool loops, return the complete assistant message, including reasoning_content and tool_calls, with matching tool-call IDs and results. When changing model or key settings, start a fresh session and record cache usage again.

tool_calls

Supported Image Input Formats

The official K3.1 Pro guide accepts base64 or ms://<file-id> images, with content as an array of objects. Public image URLs are not supported. Check the selected gateway route’s upload contract before sending images.

content

Developer API Credentials and Subscriptions

Kimi membership and Kimi Code have their own credentials, allowances and client limits. Moonshot developer API usage is billed separately. Confirm route ID, streaming, tools, limits and usage records before switching production traffic.

API Key

Kimi K3.1 Pro Reasoning and Output Parameters

Thinking is always enabled. Chat reasoning_effort accepts low, high or max (default max). Use max_completion_tokens for Chat and max_output_tokens for Responses.

Kimi K3.1 Pro Web Search: Tools and Billing

Both variants expose built-in web search, billed per actual call rather than tool declaration. Responses and Messages run it server-side; Chat continues with tool results. Moonshot’s current guide says search is being updated and is not recommended for production in the near term.

Kimi K3.1 Pro vs K3: EvoLink API Pricing and Specs

EvoLink

EvoLink live account rates in USD per 1M input / output tokens. The K3 baseline is kimi-k3; first-party list prices and independent task-cost results are identified separately.

Model	Kimi K3.1 Pro	Kimi K3
Input / output	$2 / $10	$2.85 / $14.25
Context	1.05M	1.05M
Caching	5m / 1h	Automatic cache reads
Best for	Choose Pro for token-sensitive batch jobs and multi-step coding or research workflows.	Moonshot flagship reasoning model

Kimi K3.1 Pro

Input / output$2 / $10

Context1.05M

Caching5m / 1h

Best forChoose Pro for token-sensitive batch jobs and multi-step coding or research workflows.

Kimi K3

Input / output$2.85 / $14.25

Context1.05M

CachingAutomatic cache reads

Best forMoonshot flagship reasoning model

Kimi Model Family

Same API key and balance — switch tiers without changing your integration.

Kimi K3.1 Pro

Current model

Choose Pro for token-sensitive batch jobs and multi-step coding or research workflows.

Kimi K3

Moonshot flagship reasoning model

View model

Kimi K3.1 Pro Alternatives on EvoLink

Claude Opus 4.8

Premium baseline for long-running coding agents, complex review, and judgment-heavy professional workflows.

View model

Gemini 3.1 Pro

Long-context multimodal route for document-heavy work, visual evidence, and cross-provider fallback testing.

View model

Grok 4.5

xAI reasoning and tool-use route with a 500K context window, server-side search, and configurable reasoning effort.

View model

GPT-5.6

OpenAI GPT-5.6 API model.

View model

Kimi K3.1 Pro Community Feedback and Subscription Quotas

App Access and Subscription Quotas

The launch discussion includes reports of access arriving and app or coding allowances running out quickly. These are individual product experiences, not API rate limits or benchmarks. Use completed-task quality, billed tokens, time and recovery behavior for your own evaluation.

Kimi K3.1 Pro guides and related reading

Kimi K3.1 Pro release guide

Understand the changes, access channels and announced weight-release plan.

Read guide

Kimi K3.1 Pro vs K3

Decide which workloads to evaluate, migrate or keep on K3.

Read guide

Kimi K3.1 Pro API FAQ

How do I access the Kimi K3.1 Pro API?

Create an EvoLink API key and send kimi-k3.1-pro or kimi-k3.1-pro-highspeed as the model ID. Both use the same key and base URL; switch the model ID to compare variants.

Kimi K3.1 Pro vs Highspeed: which should I choose?

Moonshot positions Highspeed as the same intelligence with faster output. It has higher input, output and cache-write rates; cache-read and search rates are the same. The speed claim and practical trade-offs are explained below the evaluation table.

How is Kimi K3.1 Pro pricing calculated?

Uncached input, output including reasoning, cache reads, 5m cache writes and 1h cache writes are billed separately. Web search adds a per-call fee only when used. Select a variant in the calculator to estimate the combined cost.

Does Kimi K3.1 Pro support images and long documents?

Both variants accept text and supported images with a 1,048,576-token shared context. Large context does not guarantee perfect recall; validate retrieval and citations. See the specification and integration details for output budgets and image formats.

Which API protocols and streaming modes does Kimi K3.1 Pro support?

Use /v1/chat/completions, /v1/responses or /v1/messages. Each supports synchronous JSON and SSE streaming through EvoLink.

How are 5m and 1h cache writes billed for Kimi K3.1 Pro?

Use five mutually exclusive token counts: ordinary input excludes cache reads and both write types. Classify 5m and 1h writes using returned usage, since existing cache entries retain their original TTL.

Does Kimi K3.1 Pro support web search?

Yes, on both Pro and Highspeed. Responses uses tools with type web_search; Messages uses web_search_20250305; Chat uses the built-in $web_search function and requires tool-result continuation. Actual calls are billed separately from tokens.

Are Kimi K3.1 Pro weights available?

Moonshot announced full weights by November 4, 2026. As checked on October 24, API availability and the announced weight-release schedule are separate; verify the published weights and license before self-hosting.

Does a Kimi subscription include developer API usage?

Moonshot developer API usage is billed separately from Kimi membership and Kimi Code. EvoLink requests use an EvoLink key and account balance.

What should I test before switching from K3 to Kimi K3.1 Pro?

Use the same repository or research task, acceptance tests and tool setup. Compare completion quality, retries, billed reasoning/output tokens and end-to-end time. Independent score gains do not guarantee lower cost on every task.