---
id: "moonshotai/kimi-k2"
also: ["moonshotai/kimi-k2", "moonshotai/Kimi-K2-Instruct", "hf/moonshotai/Kimi-K2-Instruct", "requesty/moonshot/kimi-k2", "vercel/moonshotai/kimi-k2"]
cheapest_list_usd_per_m: "0.57 in, 2.3 out, at novita on hf"
cheapest_effective_usd_per_m: "0.2993 in, 2.5 out, at vertex on requesty"
context: 131072
max_output: 98304
input: ["text"]
output: ["text"]
released: "2025-07-11"
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "max_tokens", "presence_penalty", "repetition_penalty", "seed", "stop", "temperature", "tool_choice", "tools", "top_k", "top_p"]
endpoints: 6
gateways: "hf (2), vercel (2), openrouter (1), requesty (1)"
no_training: 2
trains: 0
retains_prompts: 0
---
# MoonshotAI: Kimi K2 0711

Kimi K2 Thinking is an open-source model that operates as a "thinking agent," reasoning step-by-step while using tools to achieve state-of-the-art performance on various benchmarks. It is capable of executing up to 200-300 sequential tool calls without human intervention, allowing it to solve complex problems across a wide range of tasks. The model uses Quantization-Aware Training (QAT) to support INT4 inference, which provides a roughly 2x improvement in generation speed.