MiniMax M2.1
MiniMax M2.1 is an enhanced large language model focused on multi-language programming and real-world complex tasks. It features exceptional capabilities across Rust, Java, Golang, C++, Kotlin, Objective-C, TypeScript, JavaScript and more, with industry-leading multilingual performance that outperforms Claude Sonnet 4.5 and approaches Claude Opus 4.5. M2.1 significantly strengthens native Android and iOS development, delivers enhanced design comprehension and aesthetic expression for web/app scenarios, and provides more concise responses with improved speed and reduced token consumption. It excels across various coding agent frameworks including Claude Code, Droid (Factory AI), Cline, Kilo Code, Roo Code, and BlackBox.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| AA-LCR | 62.0% | self-reported llm-stats | link → |
| AIME 2025 | 81.0% | self-reported llm-stats | link → |
| BrowseComp | 62.0% | self-reported llm-stats | link → |
| GPQA | 81.0% | self-reported llm-stats | link → |
| Humanity's Last Exam | 22.0% | self-reported llm-stats | link → |
| IFBench | 70.0% | self-reported llm-stats | link → |
| LiveCodeBench | 78.0% | self-reported llm-stats | link → |
| MMLU-Pro | 88.0% | self-reported llm-stats | link → |
| Multi-SWE-Bench | 49.4% | self-reported llm-stats | link → |
| OctoCodingBench | 26.1% | self-reported llm-stats | link → |
| SciCode | 39.0% | self-reported llm-stats | link → |
| SWE-bench Multilingual | 72.5% | self-reported llm-stats | link → |
| SWE-Bench Verified | 67.0% | self-reported llm-stats | link → |
| SWE-Perf | 3.1% | self-reported llm-stats | link → |
| SWE-Review | 8.9% | self-reported llm-stats | link → |
| SWT-Bench | 69.3% | self-reported llm-stats | link → |
| Tau2 Telecom | 87.0% | self-reported llm-stats | link → |
| Terminal-Bench | 47.9% | self-reported llm-stats | link → |
| Toolathlon | 43.5% | self-reported llm-stats | link → |
| VIBE | 88.6% | self-reported llm-stats | link → |
| VIBE Android | 89.7% | self-reported llm-stats | link → |
| VIBE Backend | 86.7% | self-reported llm-stats | link → |
| VIBE iOS | 88.0% | self-reported llm-stats | link → |
| VIBE Simulation | 87.1% | self-reported llm-stats | link → |
| VIBE Web | 91.5% | self-reported llm-stats | link → |