DeepSeek V4 Flash 0731 vs Gemini 3.6 Flash: Benchmark Comparison
DeepSeek V4 Flash 0731 has the higher score on 9 of 20 shared benchmarks; Gemini 3.6 Flash leads on 9.
The largest observed score gap is 26.73 pts on CyberBench v1.1 , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 7 of 20 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 | Gemini 3.6 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 8.33% ±2.15 | 3.33% ±1.17 | 5.00 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 30.29% ±3.19 | 25.00% ±3.01 | 5.29 pts | Reported ±1 SE ranges overlap |
| LegalBench | 77.71% ±0.50 | 86.70% ±0.41 | 8.99 pts | Reported ±1 SE ranges do not overlap |
| EMB | 56.98% ±3.39 | 65.41% ±2.65 | 8.43 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 49.52% ±0.47 | 56.30% ±0.18 | 6.78 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 62.56% ±3.14 | 49.90% ±3.30 | 12.67 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 70.69% ±0.90 | 74.86% ±0.85 | 4.17 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 41.41% ±2.15 | 53.15% ±2.16 | 11.74 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 80.36% ±1.97 | 79.66% ±1.86 | 0.70 pts | Reported ±1 SE ranges overlap |
| BioMysteryBench | 64.44% ±1.28 | 58.52% ±1.61 | 5.93 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 4.29% ±2.44 | 4.29% ±2.44 | 0.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 89.90% ±1.69 | 93.43% ±1.33 | 3.53 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 86.21% ±0.34 | 89.28% ±0.30 | 3.07 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 38.63% ±4.38 | 30.93% ±4.07 | 7.70 pts | Reported ±1 SE ranges overlap |
| IOI | 32.72% ±2.35 | 35.06% ±5.70 | 2.33 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 87.26% ±0.97 | 88.08% ±0.94 | 0.82 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 88.80% ±1.41 | 79.60% ±1.80 | 9.20 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 74.74% ±3.44 | 64.00% ±4.29 | 10.73 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 72.08% ±5.41 | 45.36% ±3.75 | 26.73 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4 Flash 0731 average | Gemini 3.6 Flash average |
|---|---|---|
| Legal | 38.78% | 38.35% |
| Finance | 59.94% | 61.61% |
| Healthcare | 60.89% | 66.41% |
| Science | 34.37% | 31.40% |
| Academic | 88.05% | 91.35% |
| Coding | 53.69% | 49.61% |
| Cyber | 72.08% | 45.36% |
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 Flash 0731 cost | Gemini 3.6 Flash cost | DeepSeek V4 Flash 0731 latency | Gemini 3.6 Flash latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.06 | $1.90 | 16m51s | 16m24s |
| Legal Research Bench | $0.28 | $0.47 | 21m07s | 2m48s |
| LegalBench | N/A | N/A | 3.95s | 2.96s |
| EMB | $0.38 | $3.89 | 26m14s | 12m06s |
| Finance Agent (v2) | $0.28 | $1.40 | 8m10s | 3m45s |
| Tax Agent Bench | $0.16 | $0.36 | 5m37s | 105.32s |
| TaxEval v2 | N/A | N/A | 72.20s | 12.34s |
| MedCode | N/A | N/A | 2m07s | 16.68s |
| MedScribe | N/A | N/A | 77.55s | 33.55s |
| BioMysteryBench | $0.04 | $1.25 | 27m03s | 12m42s |
| Terminal-Bench Science | $1.10 | $4.39 | 1h34m | 1h52m |
| GPQA Diamond | N/A | N/A | 79.96s | 18.06s |
| MMLU Pro | N/A | N/A | 17.19s | 7.30s |
| Code Migration | $3.61 | $8.93 | 2h08m | 1h30m |
| IOI | $0.37 | $8.67 | 27m43s | 48m23s |
| LiveCodeBench | N/A | N/A | 2m07s | 31.88s |
| ProgramBench | $1.29 | $5.92 | 5h39m | 47m33s |
| SWE-bench | $0.01 | $1.19 | 3m22s | 13m08s |
| Vibe Code Bench v1.1 | $0.91 | $3.04 | 37m22s | 25m23s |
| CyberBench v1.1 | $0.19 | $1.31 | 15m28s | 11m00s |
Results available only for DeepSeek V4 Flash 0731
- Vals Index
- ProofBench v1.1
- SkillsBench
- Terminal-Bench 4.0
Results available only for Gemini 3.6 Flash
- MortgageTax
- MMMU Pro
- SAGE
- Public Benefits Bench v1.1