Kimi-K2-Thinking

Model IntroductionText

Kimi-K2-Thinking: Deep thinking, complex reasoning

Kimi-K2-Thinking is part of the Kimi lineup from Moonshot, designed for production workloads that need reliable quality at scale. Deep thinking, complex reasoning.

It accepts Text inputs and produces Text outputs, with a context window of up to 262,144 input / 32,768 output tokens. The model is tuned for chat, analysis, and tool use in enterprise pipelines.

Extended chain-of-thought reasoning helps on math, logic, and multi-step planning. Use higher reasoning effort when accuracy matters more than latency.

Benchmarks

Benchmark (Area)Score*
AIME 2024 (Math)74.5% accuracy (Max)
LiveCodeBench (Coding)70.5 Pass@1 (Max)
Codeforces Rating (Coding)2350 rating (Max)
SWE-bench Verified (Coding)60.8% resolved (Max)
MMLU-Pro (Knowledge)66.4 EM (Max)
HMMT 2024 Feb (Math)73.0 Pass@1 (Max)
GPQA Diamond (Science)67.8 Pass@1 (Max)
MRCR 1M (Long Context)60.6 MMR (Max)
Terminal Bench 2.0 (Agent)43.8 Acc (Max)
Apex Shortlist (Agent)66.0 Pass@1 (Max)
Chinese-SimpleQA (Chinese)60.8 Pass@1 (Max)
SuperCLUE (Chinese)52.99 overall (Max)

API request example

Endpoint /v1/chat/completions. Replace <API_KEY> with a key from the console.

Standard OpenAI Chat Completions. Swap `model` for the target slug.

curl https://api.ravenrouter.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <API_KEY>" \
  -d '{
  "model": "Kimi-K2-Thinking",
  "messages": [
    {
      "role": "user",
      "content": "Hello! Introduce yourself in one sentence."
    }
  ],
  "max_tokens": 64
}'

More models: API Reference → Examples by model

Kimi-K2-Thinking API — $0.45/1M input, $1.80/1M output | RavenRouter AI