GPT 5.4 Nano vs MiMo V2.5: Benchmark Comparison
GPT 5.4 Nano has the higher score on 5 of 18 shared benchmarks; MiMo V2.5 leads on 13.
The largest observed score gap is 16.08 pts on Vibe Code Bench v1.1 , where MiMo V2.5 leads.
Reported ±1 standard-error ranges overlap on 7 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 5.4 Nano | MiMo V2.5 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 0.00% ±0.00 | 1.67% ±0.00 | 1.67 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 6.25% ±1.68 | 9.13% ±2.00 | 2.88 pts | Reported ±1 SE ranges overlap |
| LegalBench | 77.92% ±0.43 | 78.89% ±0.51 | 0.97 pts | Reported ±1 SE ranges do not overlap |
| EMB | 44.75% ±3.12 | 55.09% ±3.13 | 10.34 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 38.22% ±1.19 | 36.73% ±0.11 | 1.49 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 59.10% ±0.96 | 59.26% ±0.98 | 0.16 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 26.58% ±2.52 | 29.16% ±2.71 | 2.59 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 67.42% ±0.93 | 71.83% ±0.88 | 4.42 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 41.03% ±2.26 | 31.89% ±2.02 | 9.13 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 77.09% ±1.89 | 72.15% ±1.85 | 4.94 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 77.53% ±2.42 | 81.57% ±2.05 | 4.04 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 77.17% ±0.43 | 82.93% ±0.37 | 5.76 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 73.58% ±1.06 | 80.00% ±0.96 | 6.42 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 38.08% ±3.09 | 43.27% ±3.39 | 5.18 pts | Reported ±1 SE ranges overlap |
| Code Migration | 14.47% ±4.03 | 14.23% ±3.69 | 0.24 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 84.01% ±1.04 | 81.51% ±1.07 | 2.50 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 69.80% ±2.06 | 71.00% ±2.03 | 1.20 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 26.10% ±5.08 | 42.17% ±4.57 | 16.08 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | GPT 5.4 Nano average | MiMo V2.5 average |
|---|---|---|
| Legal | 28.06% | 29.90% |
| Finance | 47.21% | 50.41% |
| Healthcare | 59.06% | 52.02% |
| Academic | 76.09% | 81.50% |
| Education | 38.08% | 43.27% |
| Coding | 48.59% | 52.23% |
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 Nano cost | MiMo V2.5 cost | GPT 5.4 Nano latency | MiMo V2.5 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.18 | $0.04 | 6m03s | 6m59s |
| Legal Research Bench | $0.14 | $0.03 | 8m47s | 6m31s |
| LegalBench | N/A | N/A | 2.70s | 11.18s |
| EMB | $0.57 | $0.08 | 29m27s | 30m38s |
| Finance Agent (v2) | $0.16 | $0.09 | 5m35s | 5m39s |
| MortgageTax | N/A | N/A | 14.53s | 32.94s |
| Tax Agent Bench | $0.07 | $0.02 | 3m40s | 8m21s |
| TaxEval v2 | N/A | N/A | 15.90s | 26.30s |
| MedCode | N/A | N/A | 8.04s | 17.70s |
| MedScribe | N/A | N/A | 20.53s | 20.26s |
| GPQA Diamond | N/A | N/A | 29.50s | 89.39s |
| MMLU Pro | N/A | N/A | 8.21s | 30.25s |
| MMMU Pro | N/A | N/A | 21.91s | 42.47s |
| SAGE | N/A | N/A | 34.33s | 52.66s |
| Code Migration | $0.42 | $0.10 | 26m34s | 37m55s |
| LiveCodeBench | N/A | N/A | 60.09s | 108.61s |
| SWE-bench | $0.10 | $0.01 | 4m23s | 4m04s |
| Vibe Code Bench v1.1 | $1.28 | $0.07 | 54m36s | 26m20s |
Results available only for GPT 5.4 Nano
None.
Results available only for MiMo V2.5
- ProofBench v1.1