MiniMax-M2
MiniMax M2 is an open-source large language model by MiniMax, built for agents and coding tasks. It delivers state-of-the-art tool use, reasoning, and search performance while maintaining exceptional cost-efficiency and speed, priced at just 8% of Claude 3.5 Sonnet’s cost and running at nearly double its inference speed (≈100 TPS). Designed for end-to-end agentic workflows, it excels at long-chain tool calling across Shell, Browser, Python, and other MCP tools. While slightly behind top overseas models in programming, it ranks among the best domestic models and top five globally on the Artificial Analysis benchmark. M2 powers the MiniMax Agent platform, available in Lightning Mode for fast tasks and Pro Mode for complex multi-step reasoning, and its weights, API, and deployment guides are freely available on Hugging Face, vLLM, and SGLang.
MiniMax-M2 is MiniMax’s model designed for various computational tasks. It processes at 50.154 tokens per second and is priced at $0.3 per million input tokens and $1.2 per million output tokens, targeting professional users.
When to Use MiniMax-M2
✓ Best For
- Data analysis and processing tasks.
- Mathematical computations requiring high accuracy.
- Applications needing efficient token management.
✗ Not Ideal For
- Users requiring high coding capabilities.
- Tasks that rely heavily on general intelligence.
How MiniMax-M2 Compares
Intelligence Index · Higher is better
Benchmark Profile
Output Speed · tok/s
Math Index
Intelligence · Coding · Math
All Benchmark Scores (26)
| Benchmark | Score | |
|---|---|---|
| Intelligence Index | 36.1 | |
| Coding Index | 29.2 | |
| Math Index | 78.3 | |
| MMLU-Pro | 82% | |
| GPQA | 777% | |
| LiveCodeBench | 826% | |
| HLE | 125% | |
| SWE-Bench Verified | 69.4% | |
| SciCode | 36.1% | |
| IFBench | 72.3% | |
| LCR | 61% | |
| TerminalBench Hard | 25.8% | |
| Tau2 | 86.8% | |
| AIME 2025 | 78.3% | |
| AA-Index | 61% | |
| BrowseComp | 44% | |
| BrowseComp-zh | 48.5% | |
| Humanity's Last Exam | 12.5% | |
| IF | 72% | |
| Multi-SWE-Bench | 36.2% | |
| SWE-bench Multilingual | 56.5% | |
| Tau2 Telecom | 87% | |
| Tau-bench | 77.2% | |
| Terminal-Bench | 46.3% | |
| Arena Chat | 22.7 | |
| Arena Coding | 9.8 |
Data: Artificial Analysis · Updated: April 2, 2026
Frequently Asked Questions (15)
