MiniMax M3.1 Flash

MiniMax M3.1 Flash is an AI model that helps software agents read code, plan changes and use tools. MiniMax launched it on 19 October 2026 through its API, M Plan and MiniMax Code.

A million-token window holds the working context

MiniMax’s model documentation, checked on 20 October, lists a one-million-token context window, multimodal input and adjustable thinking depth. A token is a piece of text or another encoded input; the context holds the material available to the model during a request.

Longer context can help an agent work across a large project. Developers still need to control which files and tools it can access and check the changes it makes.

The launch price is $0.15 input and $0.60 output

MiniMax’s launch lists $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. These are the vendor’s API prices checked on 20 October 2026. Coding subscriptions and third-party hosts have their own terms.

The provider says downloadable weights will follow later. The current release gives users hosted access through the API and its coding products.

MiniMax reports its own coding and security results

MiniMax reports 82% on CyberGym and high serving throughput in its launch evaluation. Those figures describe the provider’s testing. Model settings, tools, hardware and task length affect results, so a team’s own code and permission settings should form part of an evaluation.