MiniMax M3.1 Flash

Overview. MiniMax M3.1 Flash is an open-weights multimodal model from MiniMax. It accepts text, image, and video input and produces text, with a sparse mixture-of-experts architecture of approximately 428B total parameters and 23B active per token. Developers can use it when long-context multimodal input and configurable reasoning effort are important.

Key strengths. The model supports a 1M-token context window and configurable reasoning effort levels from low through max. Its multimodal input covers text, images, and video, while its MoE design routes each token through a subset of the model's experts.

When to use it. Long-context work that combines text with images or video, and workflows where reasoning depth should be adjustable to balance quality against speed.

Limitations. Produces text only, with no audio output. Weights are released under the custom MiniMax Community License.

Sovereignty note. MiniMax is the model creator. Where the model runs and under which terms depends on the provider endpoint.