Where is the Kimi K3 documentation?

The Kimi K3 documentation on OpenRouter includes the Open-Weight vs Closed-Weight Models: What's the Difference? blog post. The quickstart covers API keys and your first request, and the API reference lists every parameter.

How much does Kimi K3 cost?

Kimi K3 costs $0.1824/M input tokens and $11.01/M output tokens, with separate rates for Cache Read at $0.18/M tokens.

What is the context length of Kimi K3?

Kimi K3 has a 1,048,576 token context window. It supports up to 1,048,576 completion tokens.

Does Kimi K3 support tool calling and structured outputs?

Yes. Kimi K3 accepts tools for function calling on 18 of the 20 providers serving it, and requests that send tools are routed to those providers. It also supports structured outputs via a JSON schema in response_format.

What inputs and outputs does Kimi K3 support?

Kimi K3 accepts text, images, and video as input and returns text.

Which providers serve Kimi K3?

Kimi K3 is served by 20 providers on OpenRouter: Wafer, Relace, inference.net, Sail Research, Makora, Chutes, io.net, DigitalOcean and 12 more. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

When was Kimi K3 released?

Kimi K3 was released on July 16, 2026.

MoonshotAI: Kimi K3

moonshotai/kimi-k3

Model weights

Compare

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.

Modalities

In / out price

$0.1824 / $11.01per 1M

Context

1.0M

Released

Jul 16, 2026

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick: Balanced (price + speed), Nitro (fastest), or Floor (cheapest).

Latency / throughput


	$0.1824	$11.0055	$0.18	1.30s	100 tok/s	99.77%
	$0.276	$9.60	$0.13	0.53s	92 tok/s	99.97%
Fast	$0.45	$18.00	$0.315	0.66s	111 tok/s	100.00%
	$1.20	$16.00	$0.40	1.00s	94 tok/s	99.89%
	$0.192	$16.00	$0.182	0.57s	94 tok/s	100.00%
Fast	$0.153	$16.00	$0.152	0.57s	91 tok/s	100.00%
Fast	$0.45	$18.00	$0.315	0.66s	106 tok/s	99.97%
	$3.00	$15.00	$0.30	0.82s	93 tok/s	99.86%
	$3.00	$15.00	$0.30	0.49s	112 tok/s	100.00%
	$3.00	$15.00	$0.30	1.33s	78 tok/s	99.79%
	$0.345	$16.58	$0.1725	2.24s	81 tok/s	99.72%
	$0.234	$15.00	$0.204	0.68s	64 tok/s	99.94%
	$0.234	$15.00	$0.204	0.76s	64 tok/s	99.97%
	$2.30	$14.00	$0.23	1.83s	60 tok/s	99.59%
	$3.00	$15.00	$0.30	3.12s	86 tok/s	99.93%
	$3.00	$15.00	$0.30	0.29s	141 tok/s	100.00%
Fast	$4.50	$22.50	$0.45	1.27s	99 tok/s	52.54%
	$2.85	$14.25	$0.285	1.52s	50 tok/s	99.97%
	$3.00	$15.00	$0.30	1.92s	86 tok/s	100.00%
	$3.00	$15.00	$0.30	2.07s	53 tok/s	98.57%
	$3.00	$15.00	--	0.97s	120 tok/s	100.00%
	$2.55	$12.95	$0.255	1.35s	38 tok/s	100.00%
	$3.00	$15.00	$0.30	3.06s	33 tok/s	99.97%
	$3.45	$17.25	$0.345	1.72s	37 tok/s	97.88%
Fast	$4.50	$22.50	$0.45	0.48s	145 tok/s	100.00%
	$4.50	$22.50	$0.45	0.66s	44 tok/s	99.89%
	$1.85	$9.50	$0.185	1.95s	24 tok/s	97.22%
	$0.345	$16.58	$0.1725	2.25s	47 tok/s	100.00%
	$3.00	$15.00	$0.30	--	--	100.00%

Input $0.1824Output $11.0055
100 tok/s99.77% uptime
Input $0.276Output $9.60
92 tok/s99.97% uptime
Input $0.45Output $18.00
Fast
111 tok/s100.00% uptime
Input $1.20Output $16.00
94 tok/s99.89% uptime
Input $0.192Output $16.00
94 tok/s100.00% uptime
Input $0.153Output $16.00
Fast
91 tok/s100.00% uptime
Input $0.45Output $18.00
Fast
106 tok/s99.97% uptime
Input $3.00Output $15.00
93 tok/s99.86% uptime
Input $3.00Output $15.00
112 tok/s100.00% uptime
Input $3.00Output $15.00
78 tok/s99.79% uptime
Input $0.345Output $16.58
81 tok/s99.72% uptime
Input $0.234Output $15.00
64 tok/s99.94% uptime
Input $0.234Output $15.00
64 tok/s99.97% uptime
Input $2.30Output $14.00
60 tok/s99.59% uptime
Input $3.00Output $15.00
86 tok/s99.93% uptime
Input $3.00Output $15.00
141 tok/s100.00% uptime
Input $4.50Output $22.50
Fast
99 tok/s52.54% uptime
Input $2.85Output $14.25
50 tok/s99.97% uptime
Input $3.00Output $15.00
86 tok/s100.00% uptime
Input $3.00Output $15.00
53 tok/s98.57% uptime
Input $3.00Output $15.00
120 tok/s100.00% uptime
Input $2.55Output $12.95
38 tok/s100.00% uptime
Input $3.00Output $15.00
33 tok/s99.97% uptime
Input $3.45Output $17.25
37 tok/s97.88% uptime
Input $4.50Output $22.50
Fast
145 tok/s100.00% uptime
Input $4.50Output $22.50
44 tok/s99.89% uptime
Input $1.85Output $9.50
24 tok/s97.22% uptime
Input $0.345Output $16.58
47 tok/s100.00% uptime
Input $3.00Output $15.00
-- tok/s100.00% uptime

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Weighted average

Input

$0.379/M tokens

Output

$14.34/M tokens

Kimi K3 — Price history

Price history

No effective price data for this selection

	$0.140	$9.60	92.9%	15.8%
Fast	$0.152	$16.00	91.1%	10.1%
	$0.180	$11.00	90.3%	4.1%
	$0.183	$16.00	91.8%	5.1%
	$0.207	$15.00	91.4%	3.9%
	$0.207	$15.00	90.1%	2.6%
	$0.215	$16.58	75.4%	0.1%
	$0.255	$16.58	52.1%	0.0%
Fast	$0.326	$18.00	91.8%	2.8%
Fast	$0.326	$18.00	91.7%	2.6%
	$0.409	$15.00	96.0%	5.3%
	$0.446	$15.00	94.6%	34.9%
	$0.498	$12.95	89.4%	0.4%
	$0.515	$16.00	85.7%	2.4%
	$0.564	$14.00	83.9%	0.2%
	$0.579	$15.00	89.7%	2.0%
	$0.628	$17.25	90.9%	1.3%
	$0.701	$15.00	85.2%	0.1%
	$0.717	$15.00	84.5%	0.7%
	$0.763	$14.25	81.4%	1.4%
	$0.988	$22.50	86.7%	1.5%
	$0.993	$15.00	74.3%	0.7%
	$1.154	$9.50	41.8%	0.0%
	$1.189	$15.00	67.1%	0.9%
Fast	$2.205	$22.50	56.7%	0.5%
	$2.300	$15.00	25.9%	0.1%
	$3.000	$15.00	39.7%	0.7%
Fast	$3.595	$22.50	22.4%	0.0%
	—	—	—	—

Effective in $0.140Effective out $9.60
Token share
15.8%
Effective in $0.152Effective out $16.00
FastToken share
10.1%
Effective in $0.180Effective out $11.00
Token share
4.1%
Effective in $0.183Effective out $16.00
Token share
5.1%
Effective in $0.207Effective out $15.00
Token share
3.9%
Effective in $0.207Effective out $15.00
Token share
2.6%
Effective in $0.215Effective out $16.58
Token share
0.1%
Effective in $0.255Effective out $16.58
Token share
0.0%
Effective in $0.326Effective out $18.00
FastToken share
2.8%
Effective in $0.326Effective out $18.00
FastToken share
2.6%
Effective in $0.409Effective out $15.00
Token share
5.3%
Effective in $0.446Effective out $15.00
Token share
34.9%
Effective in $0.498Effective out $12.95
Token share
0.4%
Effective in $0.515Effective out $16.00
Token share
2.4%
Effective in $0.564Effective out $14.00
Token share
0.2%
Effective in $0.579Effective out $15.00
Token share
2.0%
Effective in $0.628Effective out $17.25
Token share
1.3%
Effective in $0.701Effective out $15.00
Token share
0.1%
Effective in $0.717Effective out $15.00
Token share
0.7%
Effective in $0.763Effective out $14.25
Token share
1.4%
Effective in $0.988Effective out $22.50
Token share
1.5%
Effective in $0.993Effective out $15.00
Token share
0.7%
Effective in $1.154Effective out $9.50
Token share
0.0%
Effective in $1.189Effective out $15.00
Token share
0.9%
Effective in $2.205Effective out $22.50
FastToken share
0.5%
Effective in $2.300Effective out $15.00
Token share
0.1%
Effective in $3.000Effective out $15.00
Token share
0.7%
Effective in $3.595Effective out $22.50
FastToken share
0.0%
Effective in —Effective out —
Token share —

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Throughput

145tok/s

P50, best across providers

Latency

0.29s

P50, best provider

Reasoning effortLatency / throughput

AutoExacto benchmarks

GPQA DiamondTAU-BenchParasail (Fast)92.8%--Chutes92.4%--River92.4%--Baseten92.5%78.1%Modal93.5%75.7%Parasail92.1%74.3%

+21 more providers

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Uptime (3d)

100.00%

Availability (3d)

99.88%

Availability over the last 3 days

Last 72 hours

Availability 99.88%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.90%

Without Routing

97.37%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for MoonshotAI: Kimi K3 (Artificial Analysis, Design Arena and Vals AI)
Source	Benchmark	Score
Artificial Analysis	Kimi K3 (Max) Coding Index	76.2
Artificial Analysis	Kimi K3 (Max) Agentic Index	50.0
Artificial Analysis	Kimi K3 (Max) GPQA Diamond	93.5%
Artificial Analysis	Kimi K3 (Max) HLE	46.9%
Artificial Analysis	Kimi K3 (Max) AA-LCR	88.7%
Artificial Analysis	Kimi K3 (Max) τ-Bench Banking	46.0%
Artificial Analysis	Kimi K3 (Max) GDPval-AA	52.1%
Artificial Analysis	Kimi K3 (Max) CritPt	23.4%
Artificial Analysis	Kimi K3 (Max) SciCode	59.5%
Artificial Analysis	Kimi K3 (Max) Terminal-Bench 4.0	12.6%
Artificial Analysis	Kimi K3 (Max) Terminal-Bench 2.1	85.0%
Artificial Analysis	Kimi K3 (Low) Intelligence Index	17.4
Artificial Analysis	Kimi K3 (Low) Coding Index	72.0
Artificial Analysis	Kimi K3 (Low) Agentic Index	29.3
Artificial Analysis	Kimi K3 (Low) GPQA Diamond	84.2%
Artificial Analysis	Kimi K3 (Low) HLE	25.0%
Artificial Analysis	Kimi K3 (Low) AA-LCR	79.3%
Artificial Analysis	Kimi K3 (Low) τ-Bench Banking	41.6%
Artificial Analysis	Kimi K3 (Low) GDPval-AA	31.8%
Artificial Analysis	Kimi K3 (Low) CritPt	3.1%
Artificial Analysis	Kimi K3 (Low) SciCode	52.7%
Artificial Analysis	Kimi K3 (Low) Terminal-Bench 4.0	12.6%
Artificial Analysis	Kimi K3 (Low) Terminal-Bench 2.1	82.4%
Artificial Analysis	Kimi K3 (Low) AA-Omniscience Accuracy	42.4%
Artificial Analysis	Kimi K3 (Low) AA-Omniscience Non-Hallucination Rate	23.4%
Design Arena	Kimi K3 Agents Arena Agenticgamedev Elo	1198
Design Arena	Kimi K3 Agents Arena Androidnative Elo	1256
Design Arena	Kimi K3 Agents Arena Blender3d Elo	1153
Design Arena	Kimi K3 Agents Arena Codetovideo Elo	1269
Design Arena	Kimi K3 Agents Arena Full Stack Elo	1288
Design Arena	Kimi K3 Agents Arena Godotgamedev Elo	1200
Design Arena	Kimi K3 Agents Arena Htmlslides Elo	1151
Design Arena	Kimi K3 Agents Arena Mobile Apps Elo	1253
Design Arena	Kimi K3 Agents Arena Pdf Elo	1279
Design Arena	Kimi K3 Agents Arena Python-Pptxslides Elo	1223
Design Arena	Kimi K3 Agents Arena Spreadsheet Elo	1071
Design Arena	Kimi K3 Agents Arena Webapps Elo	1273
Design Arena	Kimi K3 Models Arena 3D Elo	1365
Design Arena	Kimi K3 Models Arena Code Categories Elo	1351
Design Arena	Kimi K3 Models Arena Data Visualization Elo	1350
Design Arena	Kimi K3 Models Arena Game Development Elo	1353
Design Arena	Kimi K3 Models Arena SVG Elo	1295
Design Arena	Kimi K3 Models Arena UI Component Elo	1341
Design Arena	Kimi K3 Models Arena Website Elo	1336
Vals AI	Kimi K3 Access to Justice Benchmark	68.2%
Vals AI	Kimi K3 BioMysteryBench	72.6%
Vals AI	Kimi K3 Code Migration v1.1	28.6%
Vals AI	Kimi K3 CorpFin v2	71.6%
Vals AI	Kimi K3 CyberBench v1.1	75.2%
Vals AI	Kimi K3 CyberOffense	25.0%
Vals AI	Kimi K3 EMB v1.1	71.3%
Vals AI	Kimi K3 Finance Agent v2.1	60.0%
Vals AI	Kimi K3 Vals Finance Index	58.0%
Vals AI	Kimi K3 Financial Coaching Bench	74.5%
Vals AI	Kimi K3 GPQA Diamond	92.9%
Vals AI	Kimi K3 HLAB-Vals-1.1	12.9%
Vals AI	Kimi K3 Immigration	68.3%
Vals AI	Kimi K3 Vals Inference Provider Index	64.9%
Vals AI	Kimi K3 IOI	48.9%
Vals AI	Kimi K3 LiveCodeBench	87.2%
Vals AI	Kimi K3 LegalBench	86.2%
Vals AI	Kimi K3 Vals Legal Index	71.7%
Vals AI	Kimi K3 Legal Research Bench	46.2%
Vals AI	Kimi K3 MedCode	49.4%
Vals AI	Kimi K3 MedScribe	88.0%
Vals AI	Kimi K3 MMLU Pro	88.0%
Vals AI	Kimi K3 MMMU Pro	88.2%
Vals AI	Kimi K3 MortgageTax	66.3%
Vals AI	Kimi K3 ProgramBench	2.0%
Vals AI	Kimi K3 ProofBench v1.1	87.0%
Vals AI	Kimi K3 Public Benefits Bench v1.1	68.2%
Vals AI	Kimi K3 Vals RSI Index	20.9%
Vals AI	Kimi K3 SAGE	86.8%
Vals AI	Kimi K3 SWE-bench	93.4%
Vals AI	Kimi K3 SWE-bench Multimodal	44.2%
Vals AI	Kimi K3 Tax Agent Bench	33.4%
Vals AI	Kimi K3 TaxEval v2	75.7%
Vals AI	Kimi K3 Terminal-Bench 4.0	17.2%
Vals AI	Kimi K3 Terminal-Bench-Science	1.4%
Vals AI	Kimi K3 Vals Index	53.0%
Vals AI	Kimi K3 Vals Multimodal Index	73.4%
Vals AI	Kimi K3 Vibe Code Bench 1-100	18.2%
Vals AI	Kimi K3 Vibe Code Bench v1.1	85.0%

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

There are many coding agents, but this one is yours.

256Btokens

Hermes Agent

Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.

236Btokens

Claude Code

Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

171Btokens

Codex

A coding agent that helps you build and ship with AI

130Btokens

OpenClaw

OpenClaw is an open-source AI agent that connects to your messaging apps and takes real actions on your behalf, from running commands and browsing the web to managing files and sending emails.

74.2Btokens

Rest of OpenRouter

73.3B

Top 5 Apps

7.69B

MoonshotAI: Kimi versions

Not currently available through EU in-region routing. In-region routing docs

More models from moonshotai

Kimi K3.1 Pro

Kimi K3.1 Pro is Moonshot AI's flagship open-weight multimodal reasoning model and a post-trained successor to Kimi K3, with a 1M-token context window. It is built for long-horizon coding, 3D creation, financial research, and end-to-end agentic knowledge work, with native image and video understanding.

Text1.0M context$0.2375 / $8.411

Kimi K3.1 Pro (batch)

Text1.0M context$0.70 / $6

Kimi K3 (batch)

Text1.0M context$2.28 / $11.40

Kimi K2.7 Code

MoonshotAI: Kimi K2.7 Code is a coding-focused model in Moonshot AI's Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts. It uses a native multimodal mixture-of-experts architecture that accepts text and image input, and it always operates in a thinking mode, preserving full reasoning content across multi-turn conversations. With a 256K-token context window, it targets long-horizon coding, agentic task decomposition, and multi-turn dialogue. The model activates 32B parameters out of roughly 1T total.

Text262K context$0.60 / $3

Kimi Latest

This model always redirects to the latest model in the Kimi family.

Text262K context

Kimi K2.6

Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration. It handles complex end-to-end coding tasks across Python, Rust, and Go, and can convert prompts and visual inputs into production-ready interfaces. Its agent swarm architecture scales to hundreds of parallel sub-agents for autonomous task decomposition - delivering documents, websites, and spreadsheets in a single run without human oversight.

Text262K context$0.3895 / $1.64

Kimi K2.5

Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. Built on Kimi K2 with continued pretraining over approximately 15T mixed visual and text tokens, it delivers strong performance in general reasoning, visual coding, and agentic tool-calling.

Text262K context$0.49 / $2.50

Kimi K2 Thinking

Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning. Built on the trillion-parameter Mixture-of-Experts (MoE) architecture introduced in Kimi K2, it activates 32 billion parameters per forward pass and supports 256 k-token context windows. The model is optimized for persistent step-by-step thought, dynamic tool invocation, and complex reasoning workflows that span hundreds of turns. It interleaves step-by-step reasoning with tool use, enabling autonomous research, coding, and writing that can persist for hundreds of sequential actions without drift.

It sets new open-source benchmarks on HLE, BrowseComp, SWE-Multilingual, and LiveCodeBench, while maintaining stable multi-agent behavior through 200–300 tool calls. Built on a large-scale MoE architecture with MuonClip optimization, it combines strong reasoning depth with high inference efficiency for demanding agentic and analytical tasks.

Text262K context$0.60 / $2.50

Kimi K2 0905

Kimi K2 0905 is the September update of Kimi K2 0711. It is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It supports long-context inference up to 256k tokens, extended from the previous 128k.

This update improves agentic coding with higher accuracy and better generalization across scaffolds, and enhances frontend coding with more aesthetic and functional outputs for web, 3D, and related tasks. Kimi K2 is optimized for agentic capabilities, including advanced tool use, reasoning, and code synthesis. It excels across coding (LiveCodeBench, SWE-bench), reasoning (ZebraLogic, GPQA), and tool-use (Tau2, AceBench) benchmarks. The model is trained with a novel stack incorporating the MuonClip optimizer for stable large-scale MoE training.

Text262K context$0.60 / $2.50

Kimi K2 0711

Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It is optimized for agentic capabilities, including advanced tool use, reasoning, and code synthesis. Kimi K2 excels across a broad range of benchmarks, particularly in coding (LiveCodeBench, SWE-bench), reasoning (ZebraLogic, GPQA), and tool-use (Tau2, AceBench) tasks. It supports long-context inference up to 128K tokens and is designed with a novel training stack that includes the MuonClip optimizer for stable large-scale MoE training.

Text131K context$0.57 / $2.30

Kimi Dev 72B

Kimi-Dev-72B is an open-source large language model fine-tuned for software engineering and issue resolution tasks. Based on Qwen2.5-72B, it is optimized using large-scale reinforcement learning that applies code patches in real repositories and validates them via full test suite execution—rewarding only correct, robust completions. The model achieves 60.4% on SWE-bench Verified, setting a new benchmark among open-source models for software bug fixing and code reasoning.

Text131K context

Kimi VL A3B Thinking

Kimi-VL is a lightweight Mixture-of-Experts vision-language model that activates only 2.8B parameters per step while delivering strong performance on multimodal reasoning and long-context tasks. The Kimi-VL-A3B-Thinking variant, fine-tuned with chain-of-thought and reinforcement learning, excels in math and visual reasoning benchmarks like MathVision, MMMU, and MathVista, rivaling much larger models such as Qwen2.5-VL-7B and Gemma-3-12B. It supports 128K context and high-resolution input via its MoonViT encoder.

Text131K context

Moonlight 16B A3B Instruct

Moonlight-16B-A3B-Instruct is a 16B-parameter Mixture-of-Experts (MoE) language model developed by Moonshot AI. It is optimized for instruction-following tasks with 3B activated parameters per inference. The model advances the Pareto frontier in performance per FLOP across English, coding, math, and Chinese benchmarks. It outperforms comparable models like Llama3-3B and Deepseek-v2-Lite while maintaining efficient deployment capabilities through Hugging Face integration and compatibility with popular inference engines like vLLM12.

Text8K context

MoonshotAI: Kimi K3

Providers