DeepSeek V4.1 Flash vs Hy4 Preview: Benchmark Comparison
DeepSeek V4.1 Flash has the higher score on 7 of 19 shared benchmarks; Hy4 Preview leads on 12.
The largest observed score gap is 21.00 pts on ProofBench v1.1 , where Hy4 Preview leads.
Reported ±1 standard-error ranges overlap on 12 of 19 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 V4.1 Flash | Hy4 Preview | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Finance Index | 57.08% ±1.45 | 57.51% ±1.45 | 0.43 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 41.35% ±3.42 | 45.19% ±3.46 | 3.85 pts | Reported ±1 SE ranges overlap |
| LegalBench | 83.28% ±0.46 | 83.76% ±0.41 | 0.48 pts | Reported ±1 SE ranges overlap |
| EMB v1.1 | 67.76% ±2.56 | 68.95% ±2.40 | 1.19 pts | Reported ±1 SE ranges overlap |
| Finance Agent v2.1 | 65.43% ±2.02 | 61.62% ±2.06 | 3.80 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 26.63% ±3.14 | 32.62% ±3.38 | 5.99 pts | Reported ±1 SE ranges overlap |
| MedCode | 41.17% ±2.04 | 43.25% ±2.13 | 2.07 pts | Reported ±1 SE ranges overlap |
| MedScribe | 85.50% ±1.92 | 83.60% ±2.06 | 1.90 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 54.00% ±5.01 | 75.00% ±4.35 | 21.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 67.78% ±1.11 | 69.26% ±2.59 | 1.48 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench-Science | 0.00% ±0.00 | 1.43% ±1.43 | 1.43 pts | Reported ±1 SE ranges overlap |
| Code Migration v1.1 | 58.94% ±2.66 | 45.36% ±3.59 | 13.57 pts | Reported ±1 SE ranges do not overlap |
| IOI | 40.28% ±2.56 | 59.33% ±4.60 | 19.05 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 0.00% ±0.00 | 0.50 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 19.70% ±1.75 | 8.08% ±1.34 | 11.62 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 84.74% ±2.89 | 77.48% ±4.04 | 7.26 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 73.69% ±5.48 | 65.36% ±5.55 | 8.33 pts | Reported ±1 SE ranges overlap |
| SRE Bench | 0.76% ±0.54 | 2.29% ±0.93 | 1.53 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 64.28% ±1.25 | 68.61% ±1.21 | 4.33 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4.1 Flash average | Hy4 Preview average |
|---|---|---|
| Legal | 62.31% | 64.47% |
| Finance | 53.27% | 54.40% |
| Healthcare | 63.34% | 63.42% |
| Math | 54.00% | 75.00% |
| Science | 33.89% | 35.34% |
| Coding | 40.83% | 38.05% |
| Cyber | 37.23% | 33.82% |
| Social Mobility | 64.28% | 68.61% |
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 V4.1 Flash cost | Hy4 Preview cost | DeepSeek V4.1 Flash latency | Hy4 Preview latency |
|---|---|---|---|---|
| Vals Finance Index | $0.25 | $0.68 | 11m37s | 32m15s |
| Legal Research Bench | $0.25 | $0.85 | 11m42s | 1h04m |
| LegalBench | N/A | N/A | 6.53s | 90.56s |
| EMB v1.1 | $0.23 | $0.79 | 13m00s | 30m39s |
| Finance Agent v2.1 | $0.37 | $0.67 | 16m00s | 25m03s |
| Tax Agent Bench | $0.13 | $0.59 | 5m49s | 41m04s |
| MedCode | N/A | N/A | 35.13s | 6m39s |
| MedScribe | N/A | N/A | 53.83s | 5m50s |
| ProofBench v1.1 | $0.13 | $0.34 | 9m07s | 22m54s |
| BioMysteryBench | $0.23 | $0.36 | 14m50s | 23m30s |
| Terminal-Bench-Science | $0.47 | $2.14 | 2h57m | 3h48m |
| Code Migration v1.1 | $1.09 | $3.58 | 1h54m | 1h54m |
| IOI | $0.24 | $1.35 | 18m30s | 1h01m |
| ProgramBench | $0.90 | $12.81 | 1h02m | 2h10m |
| Terminal-Bench 4.0 | $0.50 | $3.07 | 48m22s | 2h10m |
| Vibe Code Bench v1.1 | $0.41 | $2.18 | 15m28s | 41m33s |
| CyberBench v1.1 | $0.11 | $0.46 | 13m50s | 36m43s |
| SRE Bench | $0.55 | $4.90 | 1h01m | 1h40m |
| Public Benefits Bench v1.1 | $0.07 | $0.25 | 9m59s | 1h03m |
Results available only for DeepSeek V4.1 Flash
- Vals Science Index
- Vals Inference Provider Index
- Immigration
- MysteryMechanism
- SAGE
- SkillsBench
- SWE-bench Multimodal
- Access to Justice Benchmark
- Financial Coaching Bench
Results available only for Hy4 Preview
- Vals Index
- Vals Legal Index
- HLAB-Vals-1.1