Head-to-head正面对比
MiniMax M3 vs GLM 5.2
MiniMax meets Z.ai: sourced benchmark scores side by side, dimension by dimension, with prices and specs. Every conclusion below is derived from the published dataset — no opinions, no sponsorship.
MiniMax 对 Z.ai:带来源的 benchmark 成绩逐项并排,附价格与规格。下面每条结论都从公开数据集推导——没有主观意见,没有厂商赞助。
◈
The verdict, derived from data结论(全部由数据推导)
- GLM 5.2 leads overall: 86.3 vs 79.8 on the Intelligence Score.总分上 GLM 5.2 领先:86.3 对 79.8。
- MiniMax M3 is stronger in coding.MiniMax M3 在代码上更强。
- GLM 5.2 is stronger in reasoning, knowledge, math, agent, preference.GLM 5.2 在推理、知识、数学、智能体、偏好上更强。
- MiniMax M3 is 1.7× cheaper on blended price ($0.750 vs $1.30 per 1M tokens, in/out average).混合价(输入输出均值)上 MiniMax M3 便宜 1.7 倍:$0.750 对 $1.30/百万 token。
- Budget pick: MiniMax M3 delivers 92% of the score at 58% of the blended cost.预算优先选 MiniMax M3:用 58% 的混合成本拿到 92% 的分数。
● MiniMax M3 ● GLM 5.2 · missing dimensions count as 0 in the chart缺数据的维度在图上按 0 计
≡
Spec by spec逐项规格
| MiniMax M3 | GLM 5.2 | |
|---|---|---|
| Provider厂商 | MiniMax | Z.ai |
| Released发布 | 2026-05-31 | 2026-06-16 |
| Intelligence Score智能评分 | 79.8 | 86.3 |
| Benchmark coveragebenchmark 覆盖 | 100% | 88% |
| Reasoning推理 | 80.1 | 95.4 |
| Coding代码 | 82 | 73 |
| Knowledge知识 | 82.2 | 93.6 |
| Math数学 | 85.8 | 98.6 |
| Agent智能体 | 82.2 | 100 |
| Preference偏好 | 36.2 | 43 |
| API input $/1M输入价 $/百万 | $0.300 | $0.600 |
| API output $/1M输出价 $/百万 | $1.20 | $2.00 |
| Blended $/1M混合价 $/百万 | $0.750 | $1.30 |
| Value (score per $)性价比(分数/美元) | 106.4 | 66.4 |
| Context window上下文 | 1M | 1M |
| Max output最大输出 | 512K | 182K |
| Reasoning tiers推理档位 | — | none / minimal / low / medium / high / xhigh / max |
| Open weights开放权重 | Yes支持 | Yes支持 |
| CN-direct国内直连 | Yes支持 | Yes支持 |
| Free tier免费层 | No否 | No否 |
⚔
Benchmark by benchmark逐项 benchmark
| Benchmark基准 | MiniMax M3 | GLM 5.2 |
|---|---|---|
| Humanity's Last Exam | 37.1 | 54.7 |
| GPQA Diamond | 91.2 | 91.2 |
| MMLU-Pro | 78.1 | 88.9 |
| SWE-bench Verified | 80.5 | 62.1 |
| AIME 2025 | 93.3 | 98.6 |
| LiveCodeBench | 82.15 | 69.5 |
| Terminal-Bench | 66 | 81 |
| τ²-bench | 88.9 | 99.1⚡max |
| LMArena (Chatbot Arena) | 1444 | 1471 |
MiniMax M3 dossier → 档案 → · GLM 5.2 dossier → 档案 → · How we score评分方法