Hy4 previewvsGLM-5.3
Hy4 preview 与 GLM-5.3 有 11 个共同 benchmark，但两者公开成绩的测试模式不同，下方分数仅供参考，不判定胜负。
Hy4 preview
腾讯AI实验室 · 2026-08-28 · 推理大模型
GLM-5.3
智谱AI · 2026-08-14 · 推理大模型
测试模式不同的成绩
共 11 项。两款模型的公开成绩来自不同测试模式，分数不直接可比，不计入胜负和平均分差。
| 评测项 | Hy4 preview | GLM-5.3 |
|---|---|---|
| Vals Index v2.1 | 49.94Reported best (effort unspecified) | 53.51Max (task-specific tools) |
| Program Bench | 17.50Thinking High (With Tools) | 19Max (With Tools) |
| HLE | 55.40Thinking High (With Tools) | 62.50Max (With Tools) |
| PostTrain Bench | 35.60Thinking High (With Tools) | 39.80Max (With Tools) |
| AutomationBench | 32.10Thinking High (With Tools) | 48.20Max (With Tools) |
| DeepSWE | 64.30Thinking High (With Tools) | 68.96Max (With Tools) |
| NL2Repo-Bench | 58.90Thinking High (With Tools) | 58Max (With Tools) |
| SWE-Marathon | 31.90Thinking High (With Tools) | 42.50Max (With Tools) |
| CritPt | 16.90Thinking High (No Tools) | 19.10Max (No Tools) |
| Toolathlon-Verified | 74.10Thinking High (With Tools) | 73Max (With Tools) |
| CyberGym | 78.40Thinking High (With Tools) | 84.50Max (With Tools) |
规格对比
| 字段 | Hy4 preview | GLM-5.3 |
|---|---|---|
| 发布机构 | 腾讯AI实验室 | 智谱AI |
| 发布时间 | 2026-08-28 | 2026-08-14 |
| 模型类型 | 推理大模型 | 推理大模型 |
| 架构 | MoE 架构 | MoE 架构 |
| 参数规模 | 7700亿 | 7440亿 |
| 上下文长度 | 1M | 1M |
| 最大输出 | 暂无数据 | 128K |
API 调用价格
价格优先使用 DataLearner 配置的 API 记录；缺失项不做推测。
| 价格项 | Hy4 preview | GLM-5.3 |
|---|---|---|
| 文本输入 | ¥6 / 1M tokens | $1.4 / 1M tokens |
| 文本输出 | ¥18 / 1M tokens | $4.4 / 1M tokens |
| 缓存读取 | ¥0.3 / 1M tokens | $0.26 / 1M tokens |
小结
本页正文由结构化模型、价格与 benchmark 数据生成，不使用实时 LLM 撰写。