---
id: "minimax/minimax-m3"
also: ["minimax/minimax-m3-20260531", "MiniMaxAI/Minimax-M3", "hf/MiniMaxAI/MiniMax-M3", "requesty/minimax/minimax-m3", "vercel/minimax/minimax-m3"]
card_usd_per_m: "0.3 in, 1.2 out, at minimax"
cheapest_list_usd_per_m: "0.224 in, 0.88 out, at deepinfra/fp8/flex on openrouter"
cheapest_effective_usd_per_m: "0.0607 in, 0.96 out, at gmicloud/fp8 on vercel"
context: 1048576
max_output: 512000
input: ["text", "image", "video"]
output: ["text"]
released: "2026-05-31"
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_tokens", "min_p", "presence_penalty", "reasoning", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 35
gateways: "openrouter (19), vercel (7), hf (5), requesty (4)"
no_training: 24
trains: 2
retains_prompts: 7
intelligence: 18.9000
coding: 58.6000
agentic: 29.5000
---
# MiniMax: MiniMax M3

MiniMax-M3 is a native multimodal model with 512K context running \~428B parameters and \~23B activated parameters. It brings native multimodality, enabling deeper semantic fusion across text, image, and video. M3 also introduces MiniMax Sparse Attention (MSA) to improve long context efficiency, achieving frontier-level performance across long-horizon agentic benchmarks, excelling in both coding and cowork.