DeepSeek V4 Pro 0813 vs Hy4 Preview: Benchmark Comparison
DeepSeek V4 Pro 0813 has the higher score on 4 of 17 shared benchmarks; Hy4 Preview leads on 12.
The largest observed score gap is 25.00 pts on ProofBench v1.1 , where Hy4 Preview leads.
Reported ±1 standard-error ranges overlap on 11 of 17 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 Pro 0813 | Hy4 Preview | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 47.63% ±1.12 | 49.94% ±1.15 | 2.31 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 7.50% ±2.15 | 9.17% ±2.00 | 1.67 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 40.87% ±3.42 | 45.19% ±3.46 | 4.33 pts | Reported ±1 SE ranges overlap |
| LegalBench | 82.36% ±0.44 | 83.76% ±0.41 | 1.40 pts | Reported ±1 SE ranges do not overlap |
| EMB | 52.80% ±3.06 | 58.08% ±3.14 | 5.27 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 50.39% ±0.25 | 55.06% ±0.31 | 4.67 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 58.66% ±3.14 | 63.71% ±3.26 | 5.05 pts | Reported ±1 SE ranges overlap |
| MedCode | 42.47% ±2.16 | 43.25% ±2.13 | 0.78 pts | Reported ±1 SE ranges overlap |
| MedScribe | 80.17% ±2.00 | 83.60% ±2.06 | 3.43 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 50.00% ±5.03 | 75.00% ±4.35 | 25.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 4.29% ±2.44 | 1.43% ±1.43 | 2.86 pts | Reported ±1 SE ranges overlap |
| Code Migration | 41.54% ±4.30 | 47.43% ±4.27 | 5.89 pts | Reported ±1 SE ranges overlap |
| IOI | 51.61% ±2.40 | 59.33% ±4.60 | 7.72 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 |
| Terminal-Bench 4.0 | 14.14% ±2.02 | 8.08% ±1.34 | 6.06 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 82.30% ±3.15 | 77.48% ±4.04 | 4.82 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 72.86% ±5.50 | 65.36% ±5.55 | 7.50 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | DeepSeek V4 Pro 0813 average | Hy4 Preview average |
|---|---|---|
| Legal | 43.58% | 46.04% |
| Finance | 53.95% | 58.95% |
| Healthcare | 61.32% | 63.42% |
| Math | 50.00% | 75.00% |
| Science | 4.29% | 1.43% |
| Coding | 37.92% | 38.47% |
| Cyber | 72.86% | 65.36% |
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 Pro 0813 cost | Hy4 Preview cost | DeepSeek V4 Pro 0813 latency | Hy4 Preview latency |
|---|---|---|---|---|
| Vals Index | $3.35 | $1.44 | 1h04m | 1h01m |
| Harvey's Legal Agent Benchmark | $0.17 | $0.98 | 21m53s | 29m25s |
| Legal Research Bench | $1.09 | $0.85 | 42m07s | 1h04m |
| LegalBench | N/A | N/A | 20.05s | 90.56s |
| EMB | $1.08 | $0.79 | 37m27s | 30m39s |
| Finance Agent (v2) | $0.88 | $0.59 | 16m15s | 23m45s |
| Tax Agent Bench | $0.73 | $0.59 | 26m37s | 41m04s |
| MedCode | N/A | N/A | 3m01s | 6m39s |
| MedScribe | N/A | N/A | 2m36s | 5m50s |
| ProofBench v1.1 | $0.06 | $0.34 | 12m12s | 22m54s |
| Terminal-Bench Science | $4.63 | $2.14 | 2h16m | 3h48m |
| Code Migration | $18.58 | $3.41 | 3h32m | 2h13m |
| IOI | $2.15 | $1.35 | 1h07m | 1h01m |
| ProgramBench | $0.43 | $12.81 | 1h23m | 2h10m |
| Terminal-Bench 4.0 | $3.31 | $2.10 | 1h32m | 2h10m |
| Vibe Code Bench v1.1 | $0.36 | $2.18 | 1h06m | 41m33s |
| CyberBench v1.1 | $0.59 | $0.49 | 22m33s | 36m43s |
Results available only for DeepSeek V4 Pro 0813
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- LiveCodeBench
- SkillsBench
- SWE-bench
- Vibe Code Bench 1-100
Results available only for Hy4 Preview
- BioMysteryBench
- SRE Bench
- Public Benefits Bench v1.1