Hy4 previewvsGLM-5.3
Across 12 shared benchmarks, GLM-5.3 leads overall: Hy4 preview wins 2, GLM-5.3 wins 10, with 0 ties and an average score difference of -12.14.
Hy4 preview
腾讯AI实验室 · 2026-08-28 · Reasoning model
GLM-5.3
智谱AI · 2026-08-14 · Reasoning model
Hy4 preview2 wins(17%)(83%)10 winsGLM-5.3
Benchmark scores
Grouped by capability, sorted by largest gap within each. 12 shared benchmarks.
Coding and Software Engineer
GLM-5.3 4/5| Benchmark | Hy4 preview | GLM-5.3 | Diff |
|---|---|---|---|
| SWE-Marathon | 31.903 / 5Thinking High (With Tools) | 42.501 / 5Max (With Tools) | -10.60 |
| PostTrain Bench | 35.604 / 5Thinking High (With Tools) | 39.801 / 5Max (With Tools) | -4.20 |
| DeepSWE | 64.3011 / 31Thinking High (With Tools) | 66.908 / 31Max (With Tools) | -2.60 |
| Program Bench | 17.506 / 6Thinking High (With Tools) | 195 / 6Max (With Tools) | -1.50 |
| NL2Repo-Bench | 58.902 / 11Thinking High (With Tools) | 583 / 11Max (With Tools) | +0.90 |
AI Agent - Tool Usage
GLM-5.3 3/4| Benchmark | Hy4 preview | GLM-5.3 | Diff |
|---|---|---|---|
| AutomationBench | 32.103 / 10Thinking High (With Tools) | 48.202 / 10Max (With Tools) | -16.10 |
| CyberGym | 78.403 / 4Thinking High (With Tools) | 84.501 / 4Max (With Tools) | -6.10 |
| Terminal-Bench 2.1 | 85.4010 / 47Thinking High (With Tools) | 88.203 / 47Max (With Tools) | -2.80 |
| Toolathlon-Verified | 74.103 / 8Thinking High (With Tools) | 736 / 8Max (With Tools) | +1.10 |
Agent Level Benchmark
GLM-5.3 1/1| Benchmark | Hy4 preview | GLM-5.3 | Diff |
|---|---|---|---|
| Agents' Last Exam | 22.8013 / 14Thinking High (With Tools) | 28.504 / 14Max (With Tools) | -5.70 |
General Knowledge
GLM-5.3 1/1| Benchmark | Hy4 preview | GLM-5.3 | Diff |
|---|---|---|---|
| HLE | 55.4015 / 185Thinking High (With Tools) | 62.503 / 185Max (With Tools) | -7.10 |
Productivity Knowledge
GLM-5.3 1/1| Benchmark | Hy4 preview | GLM-5.3 | Diff |
|---|---|---|---|
| GDPval-AA v2 | 1,6788 / 16Thinking High (With Tools) | 1,7693 / 16Max (With Tools) | -91 |
Specs
| Field | Hy4 preview | GLM-5.3 |
|---|---|---|
| Publisher | 腾讯AI实验室 | 智谱AI |
| Release date | 2026-08-28 | 2026-08-14 |
| Model type | Reasoning model | Reasoning model |
| Architecture | MoE | MoE |
| Parameters | 770B | 744B |
| Context length | 1M | 1M |
| Max output | Not available | 128K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Hy4 preview | GLM-5.3 |
|---|---|---|
| Text input | ¥6 / 1M tokens | Not public |
| Text output | ¥18 / 1M tokens | Not public |
| Cache read | ¥0.3 / 1M tokens | Not public |
One or both models have incomplete public pricing.
Summary
- GLM-5.3leads in:Coding and Software Engineer (4/5), AI Agent - Tool Usage (3/4), Agent Level Benchmark (1/1), General Knowledge (1/1), Productivity Knowledge (1/1)
On average across the 12 shared benchmarks, GLM-5.3 scores 12.14 higher.
Largest single-benchmark gap: GDPval-AA v2 — Hy4 preview 1,678 vs GLM-5.3 1,769 (-91).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.