MiniMax M3

Name: MiniMax M3
Author: MiniMax

MiniMaxOpen WeightsPending Human Review

MiniMax M3 is a native multimodal MoE model trained jointly on text, images, and video. It adds MiniMax Sparse Attention for efficient million-token context and targets long-horizon coding and knowledge-work agents. The model supports enabled, adaptive, and disabled reasoning modes. MiniMax launched hosted API and Agent access on June 1 and later released downloadable weights. The full checkpoint has about 428B parameters with approximately 23B activated. Its MiniMax Community License permits non-commercial use and imposes additional conditions for commercial deployment.

2026-06-01

428B total, approximately 23B activated

Mixture of Experts (MoE) with MiniMax Sparse Attention and vision encoder

MiniMax Community License

Compare with other models

Specifications

Parameters: 428B total, approximately 23B activated
Architecture: Mixture of Experts (MoE) with MiniMax Sparse Attention and vision encoder
License: MiniMax Community License
Context Window: 1,000,000 tokens
Type: multimodal
Modalities: textimagevideo

Benchmark Scores

Advanced Specifications

Model Family: MiniMax M
API Access: Available
Chat Interface: Available

Capabilities & Limitations

Capabilities: code generationreasoningtool usevisual understandingmillion-token contextadaptive reasoning
Known Limitations: Full-model self-hosting requires substantial weight memory.Agent benchmark results depend on harness and tool configuration.Custom Community License conditions apply to commercial deployment.Long-context service limits and prices differ by input length.
Notable Use Cases: coding assistantssoftware engineering agentsoffice automationvisual coding agentsdocument research
Function Calling Support: Yes
Tool Use Support: Yes

Resources

Related Models

MiniMax M2.7

MiniMax

MiniMax M2.7 improves professional software engineering, office work, and character consistency, and explores model participation in its own training and harness improvement. It is a text MoE model in the M2 family, exposed through MiniMax API and Agent products. Downloadable weights were released after the initial hosted launch. The current weight license permits non-commercial use; commercial weight deployment requires prior written authorization. Agent Team workflows are orchestration around the model, not additional model modalities.

Typetext

Parameters230B total, 10B activated

2026-03-18

Open Weights

Details Compare

MiniMax M2.5

MiniMax

MiniMax M2.5 extends the M2 family through reinforcement learning in complex real-world environments. It targets software engineering, search and tool use, and office work, with improved planning before code generation and stronger professional web research. The release provides public modified-MIT weights and official API/Agent access. A high-speed hosted service variant shares the model identity and is grouped here. Reported research results using context management depend on that agent implementation.

Typetext

Parameters230B total, 10B activated

2026-02-12

Open Weights

Details Compare

MiniMax-M2-her

MiniMax

MiniMax-M2-her is MiniMax’s language-model specialization for immersive role-play and sustained multi-turn conversation. Training focuses on maintaining character and world consistency, advancing narrative structure, and adapting to users’ implicit preferences. A dedicated Text Chat API defines role-oriented inputs, text-only interaction, a 64K context, and a 2,048-token completion limit. The named model was available during January 2026 before the January 27 technical article; the precise first public day is not established here. Downloadable weights and a checkpoint-level architecture are not verified.