Kimi K3

Kimi k3 is Moonshot AI's flagship reasoning model. Designed for complex logic, the Kimi k3 API excels in mathematics and coding. With a 128k context window, it provides reliable Chain-of-Thought processing for demanding technical tasks.

Modalities

Context

API Usage Examples

curl --request POST "https://gptproto.com/v1/chat/completions" \
  --header "Authorization: Bearer $GPTPROTO_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "model": "kimi-k3",
    "messages": [
      {
        "role": "user",
        "content": "Hello"
      }
    ]
  }'

Kimi K3 pricing

Estimate a request with real work scenarios. GPTProto token pricing is 10% below official rates.

UsageQuantityRateCost

Input tokens($2.7/1M)

tokens

$2.7/1M$0.004

Output tokens($13.5/1M)

tokens

$13.5/1M$0.0108

Cache write tokens($0.27/1M)

tokens

$0.27/1M$0.0008

Cache read tokens($0.27/1M)

tokens

$0.27/1M$0.0067

Cost per request$0.0224

Requests

Top-up amount

Top-up $100 and you get:

Top-up credits with permanent validity. You will receive a total of $100.00.

Additional 10% model discount, saving $11.1037 versus direct official MoonshotAI API calls.

Related Models

All Models

Kimi K3

Current

byMoonshotAI1.05M context$2.7/M input$13.5/M output

Gemini 3.8 Flash

byGoogle1.05M context$0.9/M input$4.5/M output

Claude Fable 5.1

byClaude1M context$9/M input$45/M output

Qwen3.8 Max 0902

byQwen$1.8/M input$5.4/M output

GLM 5.3 Flash

byZ-AI1.31M context$0.15/M input$0.5/M output

DeepSeek v4 Flash Vision Exp

byDeepSeek1.05M context$0.44/M input$1.32/M output

GLM 5.3

byZ-AI1.31M context$1.26/M input$3.96/M output

Gemini 3.7 Flash

byGoogle1.05M context$0.9/M input$4.5/M output

Grok 4.6

byGrok500K context$1.2/M input$3.6/M output

Qwen3.8 Max

byQwen1M context$1.8/M input$5.4/M output

Claude Opus 5

byClaude1M context$4.5/M input$22.5/M output

Gemini 3.6 Flash

byGoogle1.05M context$0.9/M input$4.5/M output

GPT 5.6 Luna

byOpenAI1.05M context$0.16/M input$0.96/M output

GPT 5.6 Terra

byOpenAI1.05M context$1.6/M input$9.6/M output

GPT 5.6 Sol

byOpenAI1.05M context$3.2/M input$16/M output

Grok 4.5

byGrok500K context$1.2/M input$3.6/M output

Claude Sonnet 5

byClaude1M context$1.8/M input$9/M output

MiniMax M3

byMiniMax1.05M context$0.48/M input$0.96/M output

GLM 5.2

byZ-AI1.05M context$1.26/M input$3.96/M output

Qwen3.7 Max

byQwen1M context$0.36/M input$1.44/M output

DeepSeek v4 Flash

byDeepSeek1.05M context$0.44/M input$1.32/M output

DeepSeek v4 Pro

byDeepSeek1.05M context$1.32/M input$3.96/M output

Grok 4.3

byGrok1M context$0.75/M input$1.5/M output

Kimi K2.6

byMoonshotAI262K context$0.855/M input$3.6/M output

MiniMax M2.5

byMiniMax205K context$0.24/M input$0.96/M output

Kimi K2.5

byMoonshotAI262K context$0.54/M input$2.7/M output

Doubao Seed 1.6 Thinking (Build 250715)

byBytedance262K context$0.0971/M input$0.9714/M output

Doubao Seed 1.6 Thinking (Build 250615)

byBytedance262K context$0.0971/M input$0.9714/M output

Doubao Seed 1.6 Flash (Build 250615)

byBytedance262K context$0.0182/M input$0.1821/M output

Model					Input → Output
Kimi K3Current $		1.05M	$2.70 / $13.50 per 1M	$0.27 / $0.27 per 1M	Input: TextInput: ImageInput: Document Output: Text
Gemini 3.8 Flash $		1.05M	$0.90 / $4.50 per 1M	$0.60 / $0.09 per 1M	Input: TextInput: ImageInput: VideoInput: DocumentInput: Audio Output: Text
Claude Fable 5.1 $		1M	$9.00 / $45.00 per 1M	$11.25 / $0.23 per 1M	Input: TextInput: ImageInput: Document Output: Text
Qwen3.8 Max 0902 $		—	$1.80 / $5.40 per 1M	$2.25 / $0.23 per 1M	Input: TextInput: ImageInput: VideoInput: DocumentInput: Audio Output: Text
GLM 5.3 Flash $	—	1.31M	$0.15 / $0.50 per 1M	— / $0.03 per 1M	Input: TextInput: ImageInput: VideoInput: Document Output: Text
DeepSeek v4 Flash Vision Exp $	—	1.05M	$0.44 / $1.32 per 1M	— / $0.01 per 1M	Input: TextInput: Image Output: Text
GLM 5.3 $		1.31M	$1.26 / $3.96 per 1M	— / $0.23 per 1M	Input: TextInput: ImageInput: Document Output: Text
Gemini 3.7 Flash $		1.05M	$0.90 / $4.50 per 1M	$0.60 / $0.09 per 1M	Input: TextInput: ImageInput: Document Output: Text
Grok 4.6 $		500K	$1.20 / $3.60 per 1M	— / $0.30 per 1M	Input: TextInput: Image Output: Text
Qwen3.8 Max $		1M	$1.80 / $5.40 per 1M	$2.25 / $0.23 per 1M	Input: TextInput: ImageInput: VideoInput: Document Output: Text
Claude Opus 5 $		1M	$4.50 / $22.50 per 1M	$5.63 / $0.45 per 1M	Input: TextInput: ImageInput: Document Output: Text
Gemini 3.6 Flash $		1.05M	$0.90 / $4.50 per 1M	$0.60 / $0.09 per 1M	Input: TextInput: ImageInput: Document Output: Text
GPT 5.6 Luna $		1.05M	$0.16 / $0.96 per 1M	$0.20 / $0.02 per 1M	Input: TextInput: ImageInput: Document Output: Text
GPT 5.6 Terra $		1.05M	$1.60 / $9.60 per 1M	$2.00 / $0.16 per 1M	Input: TextInput: ImageInput: Document Output: Text
GPT 5.6 Sol $		1.05M	$3.20 / $16.00 per 1M	$4.00 / $0.32 per 1M	Input: TextInput: ImageInput: Document Output: Text
Grok 4.5 $		500K	$1.20 / $3.60 per 1M	$0.30 / $0.30 per 1M	Input: TextInput: Image Output: Text
Claude Sonnet 5 $		1M	$1.80 / $9.00 per 1M	$2.25 / $0.18 per 1M	Input: TextInput: Document Output: Text
MiniMax M3 $		1.05M	$0.48 / $0.96 per 1M	$0.10 / $0.10 per 1M	Input: TextInput: ImageInput: Document Output: Text
GLM 5.2 $		1.05M	$1.26 / $3.96 per 1M	$0.23 / $0.23 per 1M	Input: TextInput: ImageInput: Document Output: Text
Qwen3.7 Max $		1M	$0.36 / $1.44 per 1M	$0.07 / $0.07 per 1M	Input: TextInput: Document Output: Text
DeepSeek v4 Flash $	—	1.05M	$0.44 / $1.32 per 1M	— / $0.01 per 1M	Input: Text Output: Text
DeepSeek v4 Pro $	—	1.05M	$1.32 / $3.96 per 1M	— / $0.04 per 1M	Input: Text Output: Text
Grok 4.3 $		1M	$0.75 / $1.50 per 1M	$0.12 / $0.12 per 1M	Input: TextInput: Image Output: Text
Kimi K2.6 $		262K	$0.85 / $3.60 per 1M	$0.14 / $0.14 per 1M	Input: TextInput: Document Output: Text
MiniMax M2.5 $		205K	$0.24 / $0.96 per 1M	$0.30 / $0.02 per 1M	Input: TextInput: Document Output: Text
Kimi K2.5 $		262K	$0.54 / $2.70 per 1M	$0.09 / $0.09 per 1M	Input: TextInput: Document Output: Text
Doubao Seed 1.6 Thinking (Build 250715) $		262K	$0.10 / $0.97 per 1M	—	Input: TextInput: Image Output: Text
Doubao Seed 1.6 Thinking (Build 250615) $		262K	$0.10 / $0.97 per 1M	—	Input: TextInput: Image Output: Text
Doubao Seed 1.6 Flash (Build 250615) $		262K	$0.02 / $0.18 per 1M	—	Input: TextInput: Image Output: Text

Key Features of the Kimi k3 API

Explore the technical foundations that make Kimi k3 a leader in AI reasoning and long-context performance.

RL-Driven Reasoning

Kimi uses Reinforcement Learning to 'think' before answering, generating a verifiable Chain-of-Thought.

128k Context Window

Process massive datasets without losing information, thanks to Moonshot AI’s long-context expertise.

94.5% MATH Score

Kimi achieves elite-level performance on hard mathematics benchmarks, ensuring computational accuracy.

OpenAI Compatible

Standardized API parameters allow for a seamless transition from GPT-4o to Kimi k3 in minutes.

Kimi k3 API Frequently Asked Questions

Find technical answers regarding Kimi k3 API integration, performance benchmarks, and reasoning capabilities to optimize your development workflow.

Why does the Kimi k3 API have a response delay?

The delay occurs because Kimi generates an internal Chain-of-Thought before providing a final answer. This Reinforcement Learning process allows Kimi to verify logic and correct potential errors in real-time. For complex math or coding tasks, this extra step significantly improves accuracy compared to standard models that respond instantly. Expect a few seconds of 'thinking' time for the model to produce its high-quality, verified results.

Is the Kimi k3 API compatible with OpenAI SDKs?

Yes, the Kimi k3 API is designed to be a drop-in replacement for OpenAI workflows. You can use the official OpenAI Python or Node.js SDKs by simply updating the base URL to our endpoint and changing the model name to kimi-k3. This allows you to leverage Kimi's superior reasoning capabilities without rewriting your entire codebase or changing your existing infrastructure.

What is the context window limit for Kimi k3?

Kimi k3 supports a massive context window of 128,000 tokens. Thanks to Moonshot AI's specialized focus on long-context processing, Kimi maintains high retrieval accuracy across the entire window. This prevents the 'lost in the middle' phenomenon often seen in other models, making Kimi ideal for analyzing large document sets, full codebases, or complex legal files where every detail matters.

How does Kimi k3 perform in coding tasks?

In coding benchmarks, Kimi k3 demonstrates deep algorithmic understanding, reaching the Expert/Candidate Master level on Codeforces. With an 88.4% score on HumanEval, Kimi is highly effective at refactoring complex functions, debugging logical errors, and generating verifiable code. It simulates execution paths during its internal thinking phase to ensure the logic holds up before outputting the final snippet.

Is my data used to train the Kimi models?

No. Data processed through our enterprise Kimi k3 API is not used to train Moonshot AI models. We prioritize data privacy and security for our users. Your inputs and the model's generated reasoning sequences remain confidential within your session, ensuring that sensitive research, proprietary code, or financial data used with Kimi is protected according to enterprise-grade standards.

What are the best use cases for Kimi k3?

Kimi k3 is best suited for tasks requiring multi-step logic. This includes quantitative financial analysis, scientific hypothesis generation, and complex legal compliance mapping. It is also excellent as the 'brain' for agentic planning, where Kimi can decompose a high-level goal into a sequence of sub-tasks. Because Kimi spends time thinking, it is perfect for accuracy-critical work rather than simple, fast chat interactions.