Gemini 3.1 Pro Preview (02/26) vs GPT 5.4 Mini: Benchmark Comparison
Gemini 3.1 Pro Preview (02/26) has the higher score on 9 of 12 shared benchmarks; GPT 5.4 Mini leads on 1.
The largest observed score gap is 15.94 pts on Vibe Code Bench v1.1 , where GPT 5.4 Mini leads.
Reported ±1 standard-error ranges overlap on 4 of 12 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 | Gemini 3.1 Pro Preview (02/26) | GPT 5.4 Mini | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 20.67% ±2.81 | 12.50% ±2.30 | 8.17 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 69.40% ±0.91 | 63.51% ±0.91 | 5.88 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 9.29% ±2.09 | 8.81% ±2.04 | 0.49 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 72.88% ±0.86 | 71.22% ±0.90 | 1.66 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 95.45% ±1.05 | 83.08% ±2.46 | 12.37 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 90.99% ±0.28 | 84.55% ±0.36 | 6.43 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 88.21% ±0.78 | 79.25% ±0.97 | 8.96 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 88.48% ±0.93 | 81.47% ±1.09 | 7.02 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 78.80% ±1.83 | 73.00% ±1.99 | 5.80 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 2.52% ±0.51 | 2.52% ±1.01 | 0.00 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 32.03% ±4.34 | 47.97% ±5.61 | 15.94 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 3.1 Pro Preview (02/26) average | GPT 5.4 Mini average |
|---|---|---|
| Legal | 20.67% | 12.50% |
| Finance | 50.52% | 47.85% |
| Academic | 91.55% | 82.29% |
| Coding | 40.37% | 40.99% |
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 | Gemini 3.1 Pro Preview (02/26) cost | GPT 5.4 Mini cost | Gemini 3.1 Pro Preview (02/26) latency | GPT 5.4 Mini latency |
|---|---|---|---|---|
| Legal Research Bench | $1.10 | $1.86 | 4m02s | 37m36s |
| MortgageTax | N/A | N/A | 23.62s | 2m06s |
| Tax Agent Bench | $0.30 | $0.81 | 100.96s | 16m46s |
| TaxEval v2 | N/A | N/A | 41.93s | 44.58s |
| GPQA Diamond | N/A | N/A | 65.76s | 63.11s |
| MMLU Pro | N/A | N/A | 23.91s | 20.00s |
| MMMU Pro | N/A | N/A | 76.99s | 50.81s |
| LiveCodeBench | N/A | N/A | 88.89s | 3m09s |
| ProgramBench | N/A | N/A | 12m36s | 53m45s |
| SWE-bench | $0.78 | $0.51 | 5m12s | 5m26s |
| Terminal-Bench 4.0 | $4.11 | $1.44 | 18m44s | 29m30s |
| Vibe Code Bench v1.1 | $3.83 | $1.19 | 20m12s | 34m15s |
Results available only for Gemini 3.1 Pro Preview (02/26)
- Vals Science Index
- LegalBench
- MedCode
- MedScribe
- ProofBench v1.1
- BioMysteryBench
- MysteryMechanism
- Terminal-Bench-Science
- SAGE
- IOI
- SRE Bench
- Financial Coaching Bench
- Public Benefits Bench v1.1
Results available only for GPT 5.4 Mini
None.