Claude Sonnet 5 vs DeepSeek V4.1 Flash: Benchmark Comparison
Claude Sonnet 5 has the higher score on 8 of 15 shared benchmarks; DeepSeek V4.1 Flash leads on 7.
The largest observed score gap is 23.33 pts on SkillsBench , where DeepSeek V4.1 Flash leads.
Reported ±1 standard-error ranges overlap on 8 of 15 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 | Claude Sonnet 5 | DeepSeek V4.1 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 41.83% ±3.43 | 41.35% ±3.42 | 0.48 pts | Reported ±1 SE ranges overlap |
| LegalBench | 83.92% ±0.46 | 83.28% ±0.46 | 0.64 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 29.37% ±3.25 | 26.63% ±3.14 | 2.74 pts | Reported ±1 SE ranges overlap |
| MedCode | 47.54% ±2.27 | 41.17% ±2.04 | 6.37 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 76.05% ±3.05 | 85.50% ±1.92 | 9.45 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 77.00% ±4.23 | 54.00% ±5.01 | 23.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench-Science | 5.71% ±2.79 | 0.00% ±0.00 | 5.71 pts | Reported ±1 SE ranges do not overlap |
| IOI | 45.00% ±2.75 | 40.28% ±2.56 | 4.72 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.50% ±0.50 | 0.50 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 46.48% ±4.49 | 69.80% ±3.88 | 23.33 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 10.61% ±1.51 | 19.70% ±1.75 | 9.09 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 81.33% ±3.05 | 84.74% ±2.89 | 3.41 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 61.91% ±5.74 | 73.69% ±5.48 | 11.78 pts | Reported ±1 SE ranges do not overlap |
| SRE Bench | 0.38% ±0.38 | 0.76% ±0.54 | 0.38 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 66.03% ±1.23 | 64.28% ±1.25 | 1.76 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | Claude Sonnet 5 average | DeepSeek V4.1 Flash average |
|---|---|---|
| Legal | 62.87% | 62.31% |
| Finance | 29.37% | 26.63% |
| Healthcare | 61.80% | 63.34% |
| Math | 77.00% | 54.00% |
| Science | 5.71% | 0.00% |
| Coding | 36.68% | 43.00% |
| Cyber | 31.14% | 37.23% |
| Social Mobility | 66.03% | 64.28% |
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 | Claude Sonnet 5 cost | DeepSeek V4.1 Flash cost | Claude Sonnet 5 latency | DeepSeek V4.1 Flash latency |
|---|---|---|---|---|
| Legal Research Bench | $2.72 | $0.25 | 25m46s | 11m42s |
| LegalBench | N/A | N/A | 4.90s | 6.53s |
| Tax Agent Bench | $1.79 | $0.13 | 19m19s | 5m49s |
| MedCode | N/A | N/A | 2m15s | 35.13s |
| MedScribe | N/A | N/A | 4m12s | 53.83s |
| ProofBench v1.1 | $1.37 | $0.13 | 14m58s | 9m07s |
| Terminal-Bench-Science | $30.30 | $0.47 | 3h06m | 2h57m |
| IOI | $12.89 | $0.24 | 59m44s | 18m30s |
| ProgramBench | $24.36 | $0.90 | 1h31m | 1h02m |
| SkillsBench | $2.90 | $0.09 | 15m12s | 3m43s |
| Terminal-Bench 4.0 | $26.05 | $0.50 | 1h45m | 48m22s |
| Vibe Code Bench v1.1 | $25.39 | $0.41 | 1h07m | 15m28s |
| CyberBench v1.1 | $1.89 | $0.11 | 16m57s | 13m50s |
| SRE Bench | $18.49 | $0.55 | 3h53m | 1h01m |
| Public Benefits Bench v1.1 | $1.29 | $0.07 | 21m09s | 9m59s |
Results available only for Claude Sonnet 5
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SWE-bench
Results available only for DeepSeek V4.1 Flash
- Vals Finance Index
- Vals Science Index
- Vals Inference Provider Index
- Immigration
- EMB v1.1
- Finance Agent v2.1
- BioMysteryBench
- MysteryMechanism
- SAGE
- Code Migration v1.1
- SWE-bench Multimodal
- Access to Justice Benchmark
- Financial Coaching Bench