---
id: "moonshotai/kimi-k3"
also: ["moonshotai/kimi-k3-20260715", "moonshotai/Kimi-K3", "hf/moonshotai/Kimi-K3", "requesty/moonshot/kimi-k3", "vercel/moonshotai/kimi-k3"]
card_usd_per_m: "3.0 in, 15.0 out, at moonshotai"
cheapest_list_usd_per_m: "0.153 in, 16.0 out, at wafer/fast on openrouter"
cheapest_effective_usd_per_m: "0.1517 in, 9.5995 out, at wafer/us on openrouter"
context: 1048576
max_output: 943718
input: ["text", "image", "video"]
output: ["text"]
released: "2026-07-16"
reasoning: ["max", "high", "low", "none"]
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_completion_tokens", "max_tokens", "min_p", "parallel_tool_calls", "presence_penalty", "reasoning", "reasoning_effort", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 73
gateways: "openrouter (37), vercel (24), requesty (7), hf (5)"
no_training: 58
trains: 5
retains_prompts: 10
intelligence: 17.4000
coding: 72
agentic: 29.3000
---
# MoonshotAI: Kimi K3

Kimi’s most capable flagship model to date, with 2.8 trillion parameters. It is built on Kimi Delta Attention (KDA), a hybrid linear attention mechanism, and Attention Residuals, with native visual understanding and a 1M-token context window. It is the world’s first open-source model in the 3-trillion-parameter class, designed for frontier intelligence scenarios including long-horizon coding, knowledge work, and reasoning.