LLMOpen weights
MiniMax-M2
MiniMax
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Specifications
- Type
- LLM
- Access
- Open weights
- Released
- October 27, 2025
- License
- MIT
- Parameters
- 230B (10B active)
- Context window
- 204,800 tokens
- Output speed
- 91 tok/s
- Latency (TTFT)
- 1.19s
- Input
- Text
- Output
- Text
- API pricing
- $0.26 in · $1.00 out / 1M tokens
Capabilities
Function callingStructured output
Benchmarks
Reasoning
| GPQA Diamond | 77.7 |
Coding
| SWE-bench Verified | 69.4 |
| Terminal-Bench | 46.3 |
| LiveCodeBench | 82.6 |
Math
| AIME 2025 | 78.3 |
General
| GAIA | 75.7 |
| MMLU-Pro | 82 |
| Humanity’s Last Exam | 12.5 |
Other
| GAIA | 75.7 |
Our take
Smaller, cheaper MoE (230B/10B) tuned for coding and agentic workflows, with strong price/performance among open models.