GPT-5.6 Luna vs GLM 5.3: Benchmark Comparison
GPT-5.6 Luna has the higher score on 6 of 21 shared benchmarks; GLM 5.3 leads on 15.
The largest observed score gap is 27.27 pts on Terminal-Bench 4.0 , where GLM 5.3 leads.
Reported ±1 standard-error ranges overlap on 8 of 21 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.
Shared benchmark results
Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.
| Benchmark | GPT-5.6 Luna | GLM 5.3 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 36.54% ±3.35 | 49.04% ±3.48 | 12.50 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 84.03% ±0.42 | 84.84% ±0.40 | 0.80 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 24.82% ±3.11 | 39.70% ±3.52 | 14.88 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 76.17% ±0.84 | 72.36% ±0.88 | 3.80 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 42.39% ±2.27 | 42.86% ±2.11 | 0.47 pts | Reported ±1 SE ranges overlap |
| MedScribe | 84.39% ±2.58 | 88.81% ±2.00 | 4.42 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 60.00% ±4.92 | 49.00% ±5.02 | 11.00 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 14.41% ±2.36 | 22.97% ±2.83 | 8.56 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench-Science | 0.00% ±0.00 | 5.71% ±2.79 | 5.71 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 91.67% ±1.74 | 88.13% ±1.85 | 3.53 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 86.04% ±0.35 | 86.77% ±0.34 | 0.73 pts | Reported ±1 SE ranges do not overlap |
| IOI | 61.78% ±11.61 | 68.44% ±7.48 | 6.67 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 1.50% ±0.86 | 1.50 pts | Reported ±1 SE ranges do not overlap |
| SkillsBench | 60.45% ±4.75 | 47.51% ±4.52 | 12.94 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 93.00% ±1.14 | 95.40% ±0.94 | 2.40 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 11.62% ±1.01 | 38.89% ±1.82 | 27.27 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 22.59% ±4.00 | 19.99% ±3.81 | 2.60 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 77.06% ±3.07 | 78.13% ±3.94 | 1.07 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 73.63% ±5.56 | 72.08% ±5.41 | 1.55 pts | Reported ±1 SE ranges overlap |
| SRE Bench | 1.53% ±0.76 | 9.16% ±1.79 | 7.63 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 61.16% ±1.27 | 68.54% ±1.21 | 7.38 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | GPT-5.6 Luna average | GLM 5.3 average |
|---|---|---|
| Legal | 60.29% | 66.94% |
| Finance | 50.49% | 56.03% |
| Healthcare | 63.39% | 65.84% |
| Math | 60.00% | 49.00% |
| Science | 7.21% | 14.34% |
| Academic | 88.85% | 87.45% |
| Coding | 46.64% | 49.98% |
| Cyber | 37.58% | 40.62% |
| Social Mobility | 61.16% | 68.54% |
Cost and latency
Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.
| Benchmark | GPT-5.6 Luna cost | GLM 5.3 cost | GPT-5.6 Luna latency | GLM 5.3 latency |
|---|---|---|---|---|
| Legal Research Bench | $0.85 | $2.24 | 39m34s | 51m50s |
| LegalBench | N/A | N/A | 5.89s | 22.48s |
| Tax Agent Bench | $0.47 | $1.80 | 42m46s | 50m53s |
| TaxEval v2 | N/A | N/A | 76.52s | 98.32s |
| MedCode | N/A | N/A | 81.28s | 2m49s |
| MedScribe | N/A | N/A | 116.89s | 2m02s |
| ProofBench v1.1 | $0.12 | $2.08 | 8m34s | 42m23s |
| MysteryMechanism | $0.11 | $1.53 | 8m13s | 30m42s |
| Terminal-Bench-Science | $0.58 | $15.63 | 2h03m | 2h45m |
| GPQA Diamond | N/A | N/A | 53.52s | 3m31s |
| MMLU Pro | N/A | N/A | 16.95s | 61.78s |
| IOI | $0.58 | $7.67 | 1h10m | 1h46m |
| ProgramBench | $0.94 | $21.96 | 41m34s | 4h16m |
| SkillsBench | $0.21 | $0.71 | 5m20s | 13m07s |
| SWE-bench | $0.04 | $0.34 | 3m21s | 14m01s |
| Terminal-Bench 4.0 | $0.73 | $9.37 | 35m55s | 1h23m |
| Vibe Code Bench 1-100 | $4.91 | $8.24 | 2h03m | 1h26m |
| Vibe Code Bench v1.1 | $0.73 | $12.45 | 25m52s | 1h04m |
| CyberBench v1.1 | $0.40 | $2.69 | 16m15s | 30m18s |
| SRE Bench | $3.44 | $27.47 | 1h42m | 2h48m |
| Public Benefits Bench v1.1 | $0.27 | $0.95 | 38m40s | 35m51s |
Results available only for GPT-5.6 Luna
- Vals Science Index
- MortgageTax
- BioMysteryBench
- MMMU Pro
- SAGE
Results available only for GLM 5.3
- Vals Finance Index
- Vals Index
- Vals Legal Index
- Vals RSI Index
- HLAB-Vals-1.1
- EMB v1.1
- Finance Agent v2.1
- Code Migration v1.1
- LiveCodeBench