MiniMaxDedicated capacity

MiniMax M3.1 Flash on capacity reserved for you

MiniMax M3.1 Flash is an open-weight mixture-of-experts model with 427B parameters, a context of 1M tokens, and text and image input. For EU sovereign processing, it can run on purpose-built AI chips reserved for you in our data center in Munich, or on-premises in your own data center. Lead time on request. Pay-per-token, it is available in our Global Model Catalog, processed outside the EU.

427Bparameters, 23B active per tokenmodel card
1Mtokens of contextmodel card
EUon dedicated capacity in our data center in Munichour promise
Openweights, under the MiniMax Community Licencelicence

What MiniMax M3.1 Flash is built for

A large open-weight model for agents that work through long chains of steps and tool calls: coding, computer use and business workflows. It reads text and images and keeps a context of 1M tokens.

Agentic workflows

Long task sequences with many tool calls, across the systems your teams work in, from sales and customer service to finance and HR. MiniMax built it to stay reliable over many steps.

Software development

Whole development workflows: building a feature, fixing an issue and testing the result, in coding agents and IDE tools.

Computer use

It reads screenshots, scanned pages and charts together with your text, so an agent can work in real applications, not only through APIs.

1M tokens of context

Long documents, a whole codebase or a long agent history fit in one request.

Reasoning you can tune

The model always thinks before it answers. The reasoning_effort parameter sets how deep it thinks.

Built with MiniMax M3.1-Flash

Why MiniMax M3.1 Flash runs on dedicated capacity

We do not serve MiniMax M3.1 Flash on shared infrastructure in Munich. On dedicated capacity it runs for you alone, sized to your workload. If it must run in your own data center, we deploy it on-premises.

Guaranteed throughput

No other tenants and no shared queue. Your requests get the capacity you reserved.

Run by us

We deploy, operate and monitor the model for you, with ISO 27001 certified operations.

Predictable monthly cost

A flat monthly cost for the capacity reserved for you, not a bill per token.

How Dedicated Capacity works

Questions about MiniMax M3.1 Flash

How fast is MiniMax M3.1 Flash?

In our Global Model Catalog, where it is processed outside the EU, we measured 158 tokens per second and 517 ms to the first token (November 2026, 10K input / 1K output, single request, server-side median). On dedicated capacity the speed depends on the sizing, so we measure it against your own prompt lengths and output ratio.

How we measure

How large is MiniMax M3.1 Flash, and what input does it take?

427B parameters in a mixture of experts, with 23B of them active per token, and a context window of 1M tokens. It takes text and images.

Model card on Hugging Face

Can I use MiniMax M3.1 Flash pay-per-token?

Yes, in our Global Model Catalog under the model ID MiniMax-M3.1-Flash. There it is processed outside the EU. For EU sovereign processing, choose dedicated capacity in Munich or an on-premises deployment.

See the Inference API

Which licence applies to MiniMax M3.1 Flash?

The MiniMax Community Licence. It allows commercial use, including hosted APIs, and asks for the line "Built with MiniMax M3.1-Flash" on a related website - which is why it is on this page. Military use is not allowed.

Read the licence

How long until MiniMax M3.1 Flash runs for us?

Lead time on request. It depends on the capacity your workload needs - tell us about it and we'll give you a date.

Talk to us

Does my data stay in the EU?

Only on dedicated capacity or on-premises. Pay-per-token requests in our Global Model Catalog are processed outside the EU. On dedicated capacity, MiniMax M3.1 Flash runs on our own hardware in our data center in Munich, and prompts and outputs are not persisted. SambaNova, the US company that makes our purpose-built AI chips, operates the software stack with remote access; our data processing agreement names it.

How EU sovereignty works here

Can we test MiniMax M3.1 Flash before we commit?

Yes. We can arrange a time-limited test on dedicated capacity before you sign, so you see how the model handles your own workload.

Ask for a test

Ready to size your dedicated capacity?

Tell us your workload and we'll put real numbers on the table quickly.