← comparisons / glm-5.1-vs-gpt-5.6-luna
GLM-5.1 vs GPT-5.6 Luna
01Head to head
| GLM-5.1 | GPT-5.6 Luna | |
|---|---|---|
| Creator | Zhipuai | OpenAI |
| Released | 2026-04-07 | 2026-07-09 |
| Context | 200K | 1.1M |
| Max output | 131K | 128K |
| Input $/M | $0.98/M | — |
| Output $/M | $3.08/M | — |
| Knowledge cutoff | — | 2026-02 |
| Modalities | text → text | text + image → text |
| Capabilities | tool-use, extended-thinking, structured-output | tool-use, extended-thinking, structured-output |
| Providers | OpenRouter | OpenAI API |
02Benchmarks · vendor-reported unless verified
| GLM-5.1 | GPT-5.6 Luna | |
|---|---|---|
| Artificial Analysis Coding Agent Index (average pass@1) | 52.7 | — |
| BrowseComp (accuracy) | — | 83.3 |
| DeepSWE | — | 67.2% |
| FrontierMath (accuracy) | — | 78.6 |
| GPQA Diamond (accuracy) | — | 92.3 |
| SWE-Atlas Codebase QnA (pass@1) | 73.2 | — |
| SWE-bench Pro | — | 62.7% |
| SWE-bench Pro (pass@1) | 19.8 | — |
| Terminal-Bench (pass@1) | 65.1 | — |
| Terminal-Bench (success rate) | — | 84.7 |