Where is the Kimi K2 Thinking documentation?

The Kimi K2 Thinking documentation on OpenRouter includes the Use OpenRouter Models in Cursor: Try it with Moonshot AI's Kimi K2 blog post. The quickstart covers API keys and your first request, and the API reference lists every parameter.

How much does Kimi K2 Thinking cost?

Kimi K2 Thinking costs $0.60/M input tokens and $2.50/M output tokens, with separate rates for Cache Read at $0.15/M tokens.

What is the context length of Kimi K2 Thinking?

Kimi K2 Thinking has a 262,144 token context window. It supports up to 98,304 completion tokens.

Does Kimi K2 Thinking support tool calling and structured outputs?

Yes. Kimi K2 Thinking accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.

Which providers serve Kimi K2 Thinking?

Kimi K2 Thinking is served by 1 provider on OpenRouter: NovitaAI.

What other text models does MoonshotAI have?

Kimi K3.1 Pro, Kimi K3, Kimi K2.7 Code and 4 more are other text models from MoonshotAI.

When was Kimi K2 Thinking released?

Kimi K2 Thinking was released on November 6, 2025.

MoonshotAI: Kimi K2 Thinking

moonshotai/kimi-k2-thinking

Model weights

Compare

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.

Modalities

In / out price

$0.60 / $2.50per 1M

Context

262K

Released

Nov 6, 2025

Providers

This model is hosted by one provider. OpenRouter forwards every request without routing.

Latency / throughput


	$0.60	$2.50	$0.15	1.42s	37 tok/s	100.00%

Input $0.60Output $2.50
37 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

$0.446/M tokens

Output

$2.50/M tokens

Kimi K2 Thinking — Price history

Price history

No effective price data for this selection

		$0.446	$2.50	34.2%	100.0%

Effective in $0.446Effective out $2.50
Token share
100.0%

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

99.84%

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 K2 Thinking (Artificial Analysis, Design Arena and Vals AI)
Source	Benchmark	Score
Artificial Analysis	K2 Think V2 Coding Index	21.0
Artificial Analysis	K2 Think V2 GPQA Diamond	71.3%
Artificial Analysis	K2 Think V2 HLE	10.1%
Artificial Analysis	K2 Think V2 IFBench	62.8%
Artificial Analysis	K2 Think V2 τ²-Bench Telecom	25.4%
Artificial Analysis	K2 Think V2 AA-LCR	57.0%
Artificial Analysis	K2 Think V2 GDPval-AA	0.0%
Artificial Analysis	K2 Think V2 CritPt	0.0%
Artificial Analysis	K2 Think V2 Terminal-Bench Hard	6.8%
Artificial Analysis	K2 Think V2 Terminal-Bench 2.1	15.0%
Artificial Analysis	Kimi K2 Thinking GPQA Diamond	83.8%
Artificial Analysis	Kimi K2 Thinking HLE	23.8%
Artificial Analysis	Kimi K2 Thinking IFBench	68.1%
Artificial Analysis	Kimi K2 Thinking τ²-Bench Telecom	93.0%
Artificial Analysis	Kimi K2 Thinking AA-LCR	72.0%
Artificial Analysis	Kimi K2 Thinking CritPt	2.6%
Artificial Analysis	Kimi K2 Thinking LiveCodeBench	85.3%
Artificial Analysis	Kimi K2 Thinking Terminal-Bench Hard	31.1%
Artificial Analysis	Kimi K2 Thinking MMLU-Pro	84.8%
Artificial Analysis	Kimi K2 Thinking AIME 2025	94.7%
Design Arena	Kimi K2 Thinking Models Arena Website Elo	1105
Vals AI	Kimi K2 Thinking AIME	85.4%
Vals AI	Kimi K2 Thinking CaseLaw v2	65.7%
Vals AI	Kimi K2 Thinking CorpFin v2	60.6%
Vals AI	Kimi K2 Thinking Finance Agent v1.1	36.6%
Vals AI	Kimi K2 Thinking GPQA Diamond	78.5%
Vals AI	Kimi K2 Thinking LiveCodeBench	63.1%
Vals AI	Kimi K2 Thinking LegalBench	80.2%
Vals AI	Kimi K2 Thinking MedQA	92.6%
Vals AI	Kimi K2 Thinking MGSM	90.1%
Vals AI	Kimi K2 Thinking MMLU Pro	81.1%
Vals AI	Kimi K2 Thinking SWE-bench	60.2%
Vals AI	Kimi K2 Thinking TaxEval v2	71.7%
Vals AI	Kimi K2 Thinking Terminal-Bench 2.0	37.1%

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.

1.55Btokens

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.

512Mtokens

Janitor AI

Janitor AI is a chatbot platform where users create and chat with custom AI characters for interactive roleplay, storytelling, and immersive fiction.

261Mtokens

Chatbox AI

On-device AI copilot

223Mtokens

Directoria

new

111Mtokens

Rest of OpenRouter

179M

Top 5 Apps

9.88M

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

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.

Text1.0M context$0.1824 / $11.01

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 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 K2 Thinking

Providers