Moonshot AIKimi K2Jul 11, 2025

Kimi K2 Turbo Preview

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.

Context Window
131K
tokens
Max Output
tokens
Released
Jul 11, 2025
Arena Rank

Capabilities

👁Vision
🧠Reasoning
🔧Tool Calling
Prompt Caching
🖥Computer Use
🎨Image Generation

Supported Parameters

Frequency Penalty
Reduce repetition
Log Probs
Token probabilities
Max Tokens
Output length limit
min_p
Presence Penalty
Encourage new topics
Repetition Penalty
Penalize repeated tokens
Response Format
JSON mode / structured output
Seed
Deterministic outputs
Stop Sequences
Custom stop tokens
structured_outputs
Temperature
Controls randomness
Tool Choice
Control tool usage
Tool Calling
Function calling support
Top K
Top-K sampling
top_logprobs
Top P
Nucleus sampling

Pricing Comparison

RouterInput / 1MOutput / 1MCached Input / 1M
Requesty$1.20$5.00$0.30
OpenRouter$0.50$2.40
Vercel AI$0.50$2.00

Benchmarks

Aider Polyglot
Aider PolyglotAider Polyglot
20.4/100

Model IDs

Requestymoonshot/kimi-k2-turbo-preview
OpenRoutermoonshotai/kimi-k2

Tags

tool-calling
Compare with another model

Related Models