Mistral Large 4 vs Step 5 Preview: Benchmark Comparison
Mistral Large 4 has the higher score on 5 of 12 shared benchmarks; Step 5 Preview leads on 7.
The largest observed score gap is 32.00 pts on ProofBench v1.1 , where Step 5 Preview leads.
Reported ±1 standard-error ranges overlap on 7 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 | Mistral Large 4 | Step 5 Preview | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Finance Index | 56.90% ±1.42 | 55.43% ±1.44 | 1.47 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 31.73% ±3.23 | 40.38% ±3.41 | 8.65 pts | Reported ±1 SE ranges do not overlap |
| EMB v1.1 | 68.97% ±2.34 | 69.55% ±2.49 | 0.58 pts | Reported ±1 SE ranges overlap |
| Finance Agent v2.1 | 61.38% ±2.02 | 58.90% ±2.05 | 2.48 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 30.39% ±3.31 | 27.27% ±3.11 | 3.12 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 10.00% ±3.02 | 42.00% ±4.96 | 32.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 67.04% ±2.67 | 67.41% ±0.37 | 0.37 pts | Reported ±1 SE ranges overlap |
| MysteryMechanism | 12.16% ±2.20 | 18.47% ±2.61 | 6.31 pts | Reported ±1 SE ranges do not overlap |
| IOI | 45.28% ±3.78 | 38.06% ±6.69 | 7.22 pts | Reported ±1 SE ranges overlap |
| SWE-bench Multimodal | 33.13% ±2.15 | 40.83% ±2.25 | 7.71 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 27.78% ±1.82 | 31.82% ±2.62 | 4.04 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 78.40% ±3.55 | 70.40% ±4.27 | 8.00 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Mistral Large 4 average | Step 5 Preview average |
|---|---|---|
| Legal | 31.73% | 40.38% |
| Finance | 53.58% | 51.91% |
| Math | 10.00% | 42.00% |
| Science | 39.60% | 42.94% |
| Coding | 46.15% | 45.28% |
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 | Mistral Large 4 cost | Step 5 Preview cost | Mistral Large 4 latency | Step 5 Preview latency |
|---|---|---|---|---|
| Vals Finance Index | $2.86 | $0.82 | 51m50s | 24m42s |
| Legal Research Bench | $5.33 | $0.84 | 1h33m | 36m08s |
| EMB v1.1 | $4.84 | $1.33 | 1h22m | 43m35s |
| Finance Agent v2.1 | $1.02 | $0.58 | 16m26s | 9m03s |
| Tax Agent Bench | $2.72 | $0.55 | 57m18s | 21m29s |
| ProofBench v1.1 | $0.64 | $0.71 | 20m59s | 38m34s |
| BioMysteryBench | $2.07 | $0.49 | 25m32s | 21m35s |
| MysteryMechanism | $2.10 | $0.60 | 35m37s | 29m20s |
| IOI | $5.39 | $2.03 | 1h24m | 1h20m |
| SWE-bench Multimodal | $5.14 | $0.61 | 27m27s | 26m06s |
| Terminal-Bench 4.0 | $10.25 | $3.11 | 1h40m | 1h58m |
| Vibe Code Bench v1.1 | $39.46 | $5.86 | 2h53m | 1h31m |
Results available only for Mistral Large 4
- Vals Legal Index
- HLAB-Vals-1.1
- MedCode
- MedScribe
- Vibe Code Bench 1-100
Results available only for Step 5 Preview
- Vals Science Index
- Terminal-Bench-Science
- Code Migration v1.1
- SRE Bench
- Public Benefits Bench v1.1