DeepSeek V4 Flash 0731 vs Step 5 Preview: Benchmark Comparison
DeepSeek V4 Flash 0731 has the higher score on 4 of 8 shared benchmarks; Step 5 Preview leads on 4.
The largest observed score gap is 14.00 pts on ProofBench v1.1 , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 4 of 8 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 | DeepSeek V4 Flash 0731 | Step 5 Preview | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 30.29% ±3.19 | 40.38% ±3.41 | 10.10 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 28.07% ±3.16 | 27.27% ±3.11 | 0.80 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 56.00% ±4.99 | 42.00% ±4.96 | 14.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 64.44% ±1.28 | 67.41% ±0.37 | 2.96 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench-Science | 4.29% ±2.44 | 2.86% ±2.01 | 1.43 pts | Reported ±1 SE ranges overlap |
| IOI | 32.72% ±2.35 | 38.06% ±6.69 | 5.33 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 18.69% ±3.54 | 31.82% ±2.62 | 13.13 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 74.74% ±3.44 | 70.40% ±4.27 | 4.34 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | DeepSeek V4 Flash 0731 average | Step 5 Preview average |
|---|---|---|
| Legal | 30.29% | 40.38% |
| Finance | 28.07% | 27.27% |
| Math | 56.00% | 42.00% |
| Science | 34.37% | 35.13% |
| Coding | 42.05% | 46.76% |
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 | DeepSeek V4 Flash 0731 cost | Step 5 Preview cost | DeepSeek V4 Flash 0731 latency | Step 5 Preview latency |
|---|---|---|---|---|
| Legal Research Bench | $0.28 | $0.84 | 21m07s | 36m08s |
| Tax Agent Bench | $0.16 | $0.55 | 5m37s | 21m29s |
| ProofBench v1.1 | $0.04 | $0.71 | 15m34s | 38m34s |
| BioMysteryBench | $0.04 | $0.49 | 27m03s | 21m35s |
| Terminal-Bench-Science | $1.10 | $4.25 | 1h34m | 2h30m |
| IOI | $0.37 | $2.03 | 27m43s | 1h20m |
| Terminal-Bench 4.0 | $0.77 | $3.11 | 1h46m | 1h58m |
| Vibe Code Bench v1.1 | $0.91 | $5.86 | 37m22s | 1h31m |
Results available only for DeepSeek V4 Flash 0731
- LegalBench
- TaxEval v2
- MedCode
- MedScribe
- GPQA Diamond
- MMLU Pro
- LiveCodeBench
- ProgramBench
- SkillsBench
- SWE-bench
- CyberBench v1.1
Results available only for Step 5 Preview
- Vals Finance Index
- Vals Science Index
- EMB v1.1
- Finance Agent v2.1
- MysteryMechanism
- Code Migration v1.1
- SWE-bench Multimodal
- SRE Bench
- Public Benefits Bench v1.1