MiniMax M1

MiniMax's prior-generation reasoning model: 456B total parameters, sparse MoE with 2 experts active per token, still useful for long-context and agent-oriented baseline comparisons against the smaller, non-commercially-licensed M2/M3 lines.

MiniMax · MiniMax

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedJune 2026SourcesMiniMaxAI/MiniMax-M1-80k on Hugging Face

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Builders who specifically want a permissively-licensed (Apache 2.0) MiniMax release for commercial use, at the cost of a larger, older architecture than the M2.x/M3 lines.

Who should use it

Builders who specifically want a permissively-licensed (Apache 2.0) MiniMax release for commercial use, at the cost of a larger, older architecture than the M2.x/M3 lines.
Teams with access to hosted inference or server-class deployment paths.

Common workflows

Long-context reasoning and agentic experiments
reasoning workflows
long-context workflows
legacy workflows
moe workflows

Deployment and hardware notes

456B total parameters, 2 experts active per token. No GGUF in the official repository -- server-class multi-GPU or hosted inference only.

License and usage notes

Apache 2.0. Open weights. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

Open weights model option for MiniMax workflows.
Builders who specifically want a permissively-licensed (Apache 2.0) MiniMax release for commercial use, at the cost of a larger, older architecture than the M2.x/M3 lines.
Tracked as Legacy baseline in the OpenSourcesAI model directory.

Limitations

456B total parameters with no official or community GGUF/llama.cpp support -- server-class or hosted inference only, not a local single-GPU deployment. The 80k-token context in this entry's name is the documented, tested window; the model's own config.json lists a much larger RoPE-scaling ceiling that is not verified as practically usable and is not used here.
456B total parameters, 2 experts active per token. No GGUF in the official repository -- server-class multi-GPU or hosted inference only.
Context window and limits: 80,000 tokens (per release name, MiniMax-M1-80k).
Verify the exact model card, provider docs, license, and serving support before production use.