Head-to-head正面对比
Qwen3.7 Max vs GLM 5.2
Qwen 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.
Qwen 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)
- On overall score they are nearly tied (88.5 vs 86.3) — decide on price, context and workload mix.总分几乎打平(88.5 对 86.3)——按价格、上下文和任务结构来选。
- Qwen3.7 Max is stronger in coding.Qwen3.7 Max 在代码上更强。
- GLM 5.2 is stronger in math, agent.GLM 5.2 在数学、智能体上更强。
- GLM 5.2 is 2.3× cheaper on blended price ($1.30 vs $2.95 per 1M tokens, in/out average).混合价(输入输出均值)上 GLM 5.2 便宜 2.3 倍:$1.30 对 $2.95/百万 token。
- Budget pick: GLM 5.2 delivers 98% of the score at 44% of the blended cost.预算优先选 GLM 5.2:用 44% 的混合成本拿到 98% 的分数。
● Qwen3.7 Max ● 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
| Qwen3.7 Max | GLM 5.2 | |
|---|---|---|
| Provider厂商 | Qwen | Z.ai |
| Released发布 | 2026-05-21 | 2026-06-16 |
| Intelligence Score智能评分 | 88.5 | 86.3 |
| Benchmark coveragebenchmark 覆盖 | 88% | 88% |
| Reasoning推理 | 95 | 95.4 |
| Coding代码 | 85.7 | 73 |
| Knowledge知识 | 94.3 | 93.6 |
| Math数学 | 85.7 | 98.6 |
| Agent智能体 | 95.6 | 100 |
| Preference偏好 | 43.8 | 43 |
| API input $/1M输入价 $/百万 | $1.48 | $0.600 |
| API output $/1M输出价 $/百万 | $4.42 | $2.00 |
| Blended $/1M混合价 $/百万 | $2.95 | $1.30 |
| Value (score per $)性价比(分数/美元) | 30 | 66.4 |
| Context window上下文 | 1M | 1M |
| Max output最大输出 | 131K | 182K |
| Reasoning tiers推理档位 | thinking on / off | 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 | Qwen3.7 Max | GLM 5.2 |
|---|---|---|
| Humanity's Last Exam | 53.5 | 54.7 |
| GPQA Diamond | 92.4 | 91.2 |
| MMLU-Pro | 89.6 | 88.9 |
| SWE-bench Verified | 80.4 | 62.1 |
| AIME 2025 | 85.7 | 98.6 |
| LiveCodeBench | 91.6 | 69.5 |
| Terminal-Bench | 69.7 | 81 |
| τ²-bench | 94.7 | 99.1⚡max |
| LMArena (Chatbot Arena) | 1474 | 1471 |
Qwen3.7 Max dossier → 档案 → · GLM 5.2 dossier → 档案 → · How we score评分方法