---
id: "minimax/minimax-m3.1-flash"
also: ["minimax/minimax-m3.1-flash-20261019", "MiniMaxAI/MiniMax-M3.1-Flash", "hf/MiniMaxAI/MiniMax-M3.1-Flash", "requesty/minimax/minimax-m3.1-flash", "vercel/minimax/minimax-m3.1-flash"]
card_usd_per_m: "0.15 in, 0.6 out, at minimax"
cheapest_list_usd_per_m: "0.1 in, 0.5 out, at inference-net on vercel"
cheapest_effective_usd_per_m: "0.0177 in, 0.6 out, at minimaxi on requesty"
context: 1048576
max_output: 524288
input: ["text", "image", "video"]
output: ["text"]
released: "2026-10-19"
reasoning: ["max", "xhigh", "high", "medium", "low"]
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", "reasoning_effort", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 25
gateways: "openrouter (13), vercel (9), requesty (2), hf (1)"
no_training: 20
trains: 2
retains_prompts: 5
intelligence: 29.1000
---
# MiniMax: MiniMax M3.1 Flash

MiniMax M3.1 Flash is a fast multimodal model in the MiniMax M3 family, built for agentic cowork automation, coding and visual tasks. It scores 72.5% on AutomationBench AA (best among open models), 73.67% on DeepSWE, 91.0 on BabyVision with tools, 82.5 on CharXiv RQ, 56% on AA Omniscience and 48% on HLE. Supports image input, tool calling, structured JSON output, reasoning and prompt caching. Ideal for multi step workflows, coding agents, frontend and presentation generation, and professional analysis at low cost.