MiniMax M3
MiniMax · #8 in our ranking · Best multimodal
MiniMax M3 is an open-weight mixture-of-experts model from MiniMax, released Jun 2026, with 428B parameters (23B active per token), a 1M-token context window, and the MiniMax Community License.
MiniMax M3 trains on text, images, and video from the first step. Its MiniMax Sparse Attention gives 9x faster prefill and 15x faster decode than M2 at 1M context, and it has enabled, adaptive, and disabled reasoning modes.
MiniMax M3 specs
- Provider
- MiniMax
- Architecture
- Mixture-of-experts
- Parameters
- 428B total · 23B active
- Context
- 1M tokens | 786K words
- License
- MiniMax Community License
- Input
- Text, Image, Video
- Released
- Jun 2026
- Weights
- Hugging Face
Sourced from the model card on Hugging Face. Last checked 2026-10-01.
Highlights
- Text, image, and video input
- 9x prefill and 15x decode speedup vs M2 at 1M context
- Adaptive reasoning mode
Where it fits for payment teams
- Partner-bank RFIs
#3 pick. Multimodal with sparse attention built for million-token contexts, useful for long correspondence threads.
- Chargebacks and disputes
#1 pick. Reads text, images, and video with a 1M-token context, so a whole dispute file fits in one run.
Harnesses in Runtime
Run MiniMax M3 through any of these harnesses on Runtime.
How to run MiniMax M3
Download the weights
Pull MiniMaxAI/MiniMax-M3 from Hugging Face. Check the MiniMax Community License before you deploy.
Serve it in your cloud
Run it with vLLM or SGLang on your own GPUs, or use a managed provider that hosts the model.
Put it to work in Runtime
Point an agent or a single skill at the model through OpenCode, then compare it to your current model with evals.
MiniMax M3 FAQ
What is the context window of MiniMax M3?
+
MiniMax M3 supports a 1M-token context window (1,048,576 tokens), roughly 786K words of English text.
How many parameters does MiniMax M3 have?
+
MiniMax M3 is a mixture-of-experts model with 428B total parameters, of which 23B are active per token.
Is MiniMax M3 free for commercial use?
+
MiniMax M3 is released under the MiniMax Community License, a custom license. Commercial use may come with conditions, so read the license on Hugging Face before you deploy or fine-tune it.
What can MiniMax M3 take as input?
+
MiniMax M3 accepts text, image, video input and generates text.
Is MiniMax M3 good for partner-bank RFIs?
+
Yes. MiniMax M3 is our #3 pick for partner-bank RFIs. Multimodal with sparse attention built for million-token contexts, useful for long correspondence threads.
Is MiniMax M3 good for chargebacks and disputes?
+
Yes. MiniMax M3 is our #1 pick for chargebacks and disputes. Reads text, images, and video with a 1M-token context, so a whole dispute file fits in one run.
Can I run MiniMax M3 in my own cloud with Runtime?
+
Yes. Download the weights from Hugging Face (MiniMaxAI/MiniMax-M3) and serve them with vLLM or SGLang in your cloud, or use a managed provider that hosts the model. Runtime agents can then use MiniMax M3 through a harness like OpenCode, with your data staying in your environment.