Last updated:
MiniMax
MiniMax's flagship model family — open-weight, multimodal and aimed at agents and coding.
Model Facts
Latest model: MiniMax M3
- Provider
- MiniMax
- Released
- Jun 1, 2026
- Context window
- 1M tokens
- Weights
- Open
Benchmarks
Full Model List
| Model | Released | Status | Context | Modalities | Note |
|---|---|---|---|---|---|
| MiniMax M3 | Jun 1, 2026 | current | 1M tokens | Text, image, video | open-weight model with 1M context and native image and video understanding |
- Official model page:
- minimax.io
- Official documentation:
- platform.minimax.io/docs
- Other sources:
- MiniMax M3 launch (DataNorth)
- MiniMax M3 (Vasundhara)
What it is
MiniMax M3, released on June 1, 2026, is an open-weight model that combines a 1M-token context window, native image and video understanding, and strong coding results — 59.0% on SWE-Bench Pro, ahead of GPT-5.5 and Gemini 3.1 Pro on that test. List API pricing at launch was $0.60 per million input tokens.
Where this information comes from
Details come from launch coverage. At launch the weights were not yet published; MiniMax said they would follow on Hugging Face and GitHub, so confirm availability before relying on them. Release dates, context windows and benchmark scores are vendor-reported unless stated, and harnesses differ between labs, so treat the numbers as indicative.
Notable traits
- Low API pricing relative to Western frontier models.
- Multimodal input in an open-weight model.
- Aimed at agentic and coding workflows.