GPT 5.4 Mini vs GLM 5.3 Flash: Benchmark Comparison
GPT 5.4 Mini has the higher score on 2 of 11 shared benchmarks; GLM 5.3 Flash leads on 8.
The largest observed score gap is 34.85 pts on Terminal-Bench 4.0 , where GLM 5.3 Flash leads.
Reported ±1 standard-error ranges overlap on 3 of 11 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 5.4 Mini | GLM 5.3 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 12.50% ±2.30 | 45.19% ±3.46 | 32.69 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 8.81% ±2.04 | 31.55% ±3.35 | 22.74 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 71.22% ±0.90 | 75.59% ±0.84 | 4.38 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 83.08% ±2.46 | 86.36% ±2.13 | 3.28 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 84.55% ±0.36 | 86.06% ±0.34 | 1.50 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 79.25% ±0.97 | 86.01% ±0.83 | 6.76 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 81.47% ±1.09 | 80.51% ±1.05 | 0.95 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 73.00% ±1.99 | 92.00% ±1.21 | 19.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 2.52% ±1.01 | 37.37% ±2.67 | 34.85 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 47.97% ±5.61 | 30.76% ±5.28 | 17.21 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | GPT 5.4 Mini average | GLM 5.3 Flash average |
|---|---|---|
| Legal | 12.50% | 45.19% |
| Finance | 40.01% | 53.57% |
| Academic | 82.29% | 86.14% |
| Coding | 40.99% | 48.13% |
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 5.4 Mini cost | GLM 5.3 Flash cost | GPT 5.4 Mini latency | GLM 5.3 Flash latency |
|---|---|---|---|---|
| Legal Research Bench | $1.86 | $0.12 | 37m36s | 51m07s |
| Tax Agent Bench | $0.81 | $0.14 | 16m46s | 21m37s |
| TaxEval v2 | N/A | N/A | 44.58s | 53.06s |
| GPQA Diamond | N/A | N/A | 63.11s | 2m11s |
| MMLU Pro | N/A | N/A | 20.00s | 44.76s |
| MMMU Pro | N/A | N/A | 50.81s | 28.22s |
| LiveCodeBench | N/A | N/A | 3m09s | 5m02s |
| ProgramBench | N/A | N/A | 53m45s | 5h30m |
| SWE-bench | $0.51 | $0.02 | 5m26s | 1h04m |
| Terminal-Bench 4.0 | $1.44 | $1.22 | 29m30s | 1h01m |
| Vibe Code Bench v1.1 | $1.19 | $2.35 | 34m15s | 1h24m |
Results available only for GPT 5.4 Mini
- MortgageTax
- SAGE
Results available only for GLM 5.3 Flash
- Vals Finance Index
- Vals Index
- Vals Legal Index
- HLAB-Vals-1.1
- LegalBench
- EMB v1.1
- Finance Agent v2.1
- MedScribe
- ProofBench v1.1
- Terminal-Bench-Science
- Code Migration v1.1
- IOI
- SkillsBench
- Vibe Code Bench 1-100
- CyberBench v1.1
- SRE Bench