MiniMax M3.1 Flash open weights on BenchGecko · MiniMax Community License on Hugging Face
MiniMax published open weights for M3.1 Flash under the MiniMax Community License. BenchGecko updates the row with Hugging Face id MiniMaxAI/MiniMax-M3.1-Flash and keeps the $0.15/$0.60 per M token rates.
What changed
MiniMax published open weights for MiniMax M3.1 Flash on Hugging Face as MiniMaxAI/MiniMax-M3.1-Flash. The repository carries the MiniMax Community License. The API model has been listed on BenchGecko since 2026-10-19. BenchGecko updates the row to open weights and sets the Hugging Face id.
vLLM published day 0 support on 2026-10-23, including NVFP4 KV cache and MoE weights plus the DSpark drafter.
Specs
From the Hugging Face model card and the existing BenchGecko row:
- Parameters · about 428B total, about 23B active per token
- Context · 1M tokens, stored as 1,048,576 on BenchGecko
- Max output · 524,288 tokens on OpenRouter
- Inputs · text, image and video
- Output · text
- Reasoning · always on, with effort from low to max
- Weights · open under the MiniMax Community License
License terms
Under the MiniMax Community License, non commercial use is free. Commercial use must display Built with MiniMax M3.1-Flash. Products and services with yearly revenue above $20M need prior written authorization from MiniMax. Below that threshold, commercial use needs a one time notice to MiniMax.
Pricing on BenchGecko
First party rates are unchanged: $0.15 input, $0.60 output and $0.015 cached input per M tokens. OpenRouter lists minimax/minimax-m3.1-flash at the same rates with a 1,048,576 token context.
Vendor-reported benchmarks
MiniMax reports these results on the model card for the official weights. BenchGecko records them as vendor claims.
- Automation-Bench v1.0.6 · 62.8
- BankerToolBench · 76.2
- Terminal Bench 4.0 · 38.1
- OSWorld 2.0 · 61.7
- Agents Last Exam · 29.6
- DeepSWE v1.1 · 73.7%
- CyberGym · 82.2
Sources
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.