Kimi-K3

modified-mit

Kimi K3 is Moonshot AI's 2.8T-parameter open-weight multimodal reasoning model. Built for complex coding, knowledge work, and long-horizon agentic workflows, it excels at navigating large repositories, calling tools, debugging, and iterating on images, logs, tests, and runtime feedback. Its architecture relies on KDA and Attention Residuals for computational efficiency.

Context
1.0M
Input price
$2.85/1M tokens
Output price
$14.25/1M tokens

Capabilities

vision
tool call
structured output
reasoning
json mode
streaming
fine tuning
batch

Details

Model IDmoonshotai/Kimi-K3
Provider DeepInfra
FamilyKimi-K3
Licensemodified-mit
Parameters—
Statusactive
Input modalitiestext, image
Output modalitiestext
Architecture—
Knowledge cutoff—
Training data cutoff—
Release date—
Deprecation date—
Typechat
Reasoning tokensYes
Max input—
Open weightYes
Sourceofficial
Last updated

Tools

Function Callingfunction_calling
Call external functions and APIs

Endpoints

Chat CompletionsPOST
Generate chat responses with messageshttps://api.deepinfra.com/v1/openai/v1/chat/completions

Pricing

Input
$2.85
Output
$14.25
Cache write
—
Cache read
$0.28
Batch in
—
Batch out
—

Price across providers

ProviderInputOutputSince launch
Vercel AI Gatewaycheapest$0.18$10↓94%
DeepInfra$2.85$14.25—
Novita AI$3$15—

API

GET/v1/models/deepinfra/moonshotai/Kimi-K3

Kimi-K3

moonshotai/Kimi-K3

Kimi K3 is Moonshot AI's 2.8T-parameter open-weight multimodal reasoning model. Built for complex coding, knowledge work, and long-horizon agentic workflows, it excels at navigating large repositories, calling tools, debugging, and iterating on images, logs, tests, and runtime feedback. Its architecture relies on KDA and Attention Residuals for computational efficiency.

Changes · 3 entries
moonshotai/Kimi-K3update00710a0Oct 6, 2026, 06:38 AM
capabilitiesreasoningyes
reasoning_tokensyes
moonshotai/Kimi-K3update632cde3Oct 5, 2026, 06:54 AM
quantizationmxfp4
moonshotai/Kimi-K3create3cddff5Aug 13, 2026, 02:54 AM