GLM-5.2 vs Hy4 preview: Specs & Benchmark Comparison
GLM-5.2 is developed by Zhipu AI, while Hy4 preview comes from Tencent. GLM-5.2 was released in June 2026, and Hy4 preview followed 2 months later in August 2026. The two are similar in size, at about 753 billion and 770 billion parameters respectively.
The two models share 11 published benchmarks. Hy4 preview leads on 10 of them, GLM-5.2 on 1. The widest gaps are on Program Bench, where GLM-5.2 scores 63.7% against 17.5%; Toolathlon, where Hy4 preview scores 74.1% against 48.2%. Averaged across everything we track, GLM-5.2 sits at 60.9% and Hy4 preview at 58.7%.
GLM-5.2 is available through an API at $1.40 per million input tokens with a 1M-token context window. We do not track a hosted API for Hy4 preview, so pricing and context cannot be compared directly.
| Characteristic | GLM-5.2 | Hy4 preview |
|---|---|---|
| Company | Zhipu AI | Tencent |
| Release Date | June 16, 2026 | August 28, 2026 |
| Parameters | 753B | 770B |
| Multimodal | No | No |
| Context (input) | 1.0M | — |
| Context (output) | 131K | — |
| Input Price / 1M | $1.40 | — |
| Output Price / 1M | $4.40 | — |
| Average Score | 60.9% | 58.7% |
| Benchmarks | ||
| Program Bench | 63.7% | 17.5% |
| Toolathlon | 48.2% | 74.1% |
| SWE-Marathon | 13.0% | 31.9% |
| DeepSWE | 46.2% | 64.3% |
| NL2Repo | 48.9% | 58.9% |
| MCP Atlas | 76.8% | 83.7% |
| SWE-Bench Pro | 62.1% | 65.7% |
| Terminal-Bench 2.1 | 82.7% | 85.4% |
| GPQA | 91.0% | 92.3% |
| PostTrainBench | 34.3% | 35.6% |
| CritPT | 16.7% | 16.9% |
Visual Benchmark Comparison
Verdict
Hy4 preview leads in 1 out of 2 comparison categories.
Both models show comparable average scores: GLM-5.2 — 0.6, Hy4 preview — 0.6.
Hy4 preview is newer: released 8/28/2026 vs 6/16/2026.
More About These Models
Related Comparisons
Frequently Asked Questions
Which is better for coding — GLM-5.2 or Hy4 preview?
Which model is cheaper — GLM-5.2 or Hy4 preview?
Which has a larger context window — GLM-5.2 or Hy4 preview?
The GLM-5.2 and Hy4 preview comparison is updated for 2026. Data includes benchmark results, API pricing, context window size and other specifications. For more detailed information, visit the GLM-5.2 or Hy4 preview page. See also the complete list of AI model comparisons.