Claude Opus 4.7 vs DeepSeek V4 Flash 0731: Benchmark Comparison
Claude Opus 4.7 has the higher score on 11 of 16 shared benchmarks; DeepSeek V4 Flash 0731 leads on 4.
The largest observed score gap is 13.44 pts on MedCode , where Claude Opus 4.7 leads.
Reported ±1 standard-error ranges overlap on 7 of 16 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 Opus 4.7 | DeepSeek V4 Flash 0731 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 6.67% ±2.00 | 8.33% ±2.15 | 1.67 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 38.46% ±3.38 | 30.29% ±3.19 | 8.17 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 85.25% ±0.45 | 77.71% ±0.50 | 7.54 pts | Reported ±1 SE ranges do not overlap |
| EMB | 63.73% ±3.11 | 56.98% ±3.39 | 6.75 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 51.51% ±0.49 | 49.52% ±0.47 | 1.99 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 63.86% ±3.14 | 62.56% ±3.14 | 1.30 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 75.27% ±0.84 | 70.69% ±0.90 | 4.58 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 54.86% ±2.21 | 41.41% ±2.15 | 13.44 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 82.95% ±1.98 | 80.36% ±1.97 | 2.59 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 90.15% ±1.54 | 89.90% ±1.69 | 0.25 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 89.87% ±0.30 | 86.21% ±0.34 | 3.66 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 43.88% ±4.22 | 38.63% ±4.38 | 5.25 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 85.07% ±1.01 | 87.26% ±0.97 | 2.19 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 |
| SWE-bench | 82.00% ±1.72 | 88.80% ±1.41 | 6.80 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 71.00% ±4.51 | 74.74% ±3.44 | 3.73 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | Claude Opus 4.7 average | DeepSeek V4 Flash 0731 average |
|---|---|---|
| Legal | 43.46% | 38.78% |
| Finance | 63.59% | 59.94% |
| Healthcare | 68.91% | 60.89% |
| Academic | 90.01% | 88.05% |
| Coding | 56.39% | 57.89% |
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 Opus 4.7 cost | DeepSeek V4 Flash 0731 cost | Claude Opus 4.7 latency | DeepSeek V4 Flash 0731 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $10.81 | $0.06 | 24m44s | 16m51s |
| Legal Research Bench | $3.06 | $0.28 | 8m32s | 21m07s |
| LegalBench | N/A | N/A | 26.62s | 3.95s |
| EMB | $12.76 | $0.38 | 38m36s | 26m14s |
| Finance Agent (v2) | $4.03 | $0.28 | 6m00s | 8m10s |
| Tax Agent Bench | $1.27 | $0.16 | 4m39s | 5m37s |
| TaxEval v2 | N/A | N/A | 62.07s | 72.20s |
| MedCode | N/A | N/A | 54.25s | 2m07s |
| MedScribe | N/A | N/A | 67.50s | 77.55s |
| GPQA Diamond | N/A | N/A | 3m23s | 79.96s |
| MMLU Pro | N/A | N/A | 3m08s | 17.19s |
| Code Migration | $36.39 | $3.61 | 1h12m | 2h08m |
| LiveCodeBench | N/A | N/A | 7m42s | 2m07s |
| ProgramBench | N/A | N/A | 54m07s | 5h39m |
| SWE-bench | $2.42 | $0.01 | 7m22s | 3m22s |
| Vibe Code Bench v1.1 | $21.41 | $0.91 | 35m42s | 37m22s |
Results available only for Claude Opus 4.7
- Vals RSI Index
- MortgageTax
- MMMU Pro
- SAGE
Results available only for DeepSeek V4 Flash 0731
- Vals Index
- ProofBench v1.1
- BioMysteryBench
- Terminal-Bench Science
- IOI
- SkillsBench
- Terminal-Bench 4.0
- CyberBench v1.1