GPT 5.5 vs GLM 5.3: Benchmark Comparison
GPT 5.5 has the higher score on 9 of 19 shared benchmarks; GLM 5.3 leads on 10.
The largest observed score gap is 14.70 pts on SkillsBench , where GPT 5.5 leads.
Reported ±1 standard-error ranges overlap on 4 of 18 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 5.5 | GLM 5.3 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals RSI Index | 16.48% | 22.91% | 6.43 pts | Uncertainty comparison unavailable |
| Harvey's Legal Agent Benchmark | 3.75% ±1.17 | 8.33% ±2.00 | 4.58 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 40.38% ±3.41 | 49.04% ±3.48 | 8.65 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 86.52% ±0.41 | 84.84% ±0.40 | 1.68 pts | Reported ±1 SE ranges do not overlap |
| EMB | 64.54% ±2.87 | 56.34% ±3.32 | 8.20 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 51.76% ±0.55 | 55.84% ±2.07 | 4.08 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 60.46% ±3.22 | 73.09% ±2.97 | 12.63 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 74.98% ±0.85 | 72.36% ±0.88 | 2.62 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 49.10% ±2.19 | 42.86% ±2.11 | 6.24 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 86.87% ±1.93 | 88.81% ±2.00 | 1.94 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 93.18% ±1.29 | 88.13% ±1.85 | 5.05 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 88.14% ±0.32 | 86.77% ±0.34 | 1.37 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 45.16% ±4.16 | 44.22% ±4.29 | 0.94 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 85.30% ±1.02 | 80.53% ±1.06 | 4.76 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 1.50% ±0.86 | 1.00 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 62.21% ±4.52 | 47.51% ±4.52 | 14.70 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 82.60% ±1.70 | 95.40% ±0.94 | 12.80 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 69.85% ±4.54 | 78.13% ±3.94 | 8.28 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 60.89% ±1.27 | 68.54% ±1.21 | 7.65 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | GPT 5.5 average | GLM 5.3 average |
|---|---|---|
| Legal | 43.55% | 47.40% |
| Finance | 62.93% | 64.41% |
| Healthcare | 67.98% | 65.84% |
| Academic | 90.66% | 87.45% |
| Coding | 57.60% | 57.88% |
| Social Mobility | 60.89% | 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 5.5 cost | GLM 5.3 cost | GPT 5.5 latency | GLM 5.3 latency |
|---|---|---|---|---|
| Vals RSI Index | $857.16 | $61.84 | 90h00m | 90h00m |
| Harvey's Legal Agent Benchmark | $4.60 | $4.30 | 12m14s | 42m38s |
| Legal Research Bench | $7.40 | $2.24 | 37m45s | 51m50s |
| LegalBench | N/A | N/A | 18.14s | 22.48s |
| EMB | $3.27 | $3.79 | 15m11s | 36m12s |
| Finance Agent (v2) | $4.15 | $1.07 | 11m02s | 15m50s |
| Tax Agent Bench | $4.07 | $1.80 | 15m52s | 50m53s |
| TaxEval v2 | N/A | N/A | 95.83s | 98.32s |
| MedCode | N/A | N/A | 2m40s | 2m49s |
| MedScribe | N/A | N/A | 2m13s | 2m02s |
| GPQA Diamond | N/A | N/A | 112.33s | 3m31s |
| MMLU Pro | N/A | N/A | 42.11s | 61.78s |
| Code Migration | $6.44 | $24.91 | 31m35s | 3h47m |
| LiveCodeBench | N/A | N/A | 2m47s | 4m07s |
| ProgramBench | $6.95 | $21.96 | 22m03s | 4h16m |
| SkillsBench | $2.54 | $0.71 | 6m55s | 13m07s |
| SWE-bench | $1.36 | $0.34 | 7m06s | 14m01s |
| Vibe Code Bench v1.1 | $16.66 | $12.45 | 31m52s | 1h04m |
| Public Benefits Bench v1.1 | $3.97 | $0.95 | 39m37s | 35m51s |
Results available only for GPT 5.5
- MortgageTax
- MMMU Pro
- SAGE
- SRE Bench
- Time Horizon Index: KSP
Results available only for GLM 5.3
- Vals Index
- ProofBench v1.1
- MysteryMechanism
- Terminal-Bench Science
- IOI
- Terminal-Bench 4.0
- Vibe Code Bench 1-100
- CyberBench v1.1