DeepSeek V4 Flash 0731 vs Grok 4.5: Benchmark Comparison
DeepSeek V4 Flash 0731 has the higher score on 9 of 22 shared benchmarks; Grok 4.5 leads on 12.
The largest observed score gap is 25.00 pts on ProofBench v1.1 , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 11 of 22 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 | Grok 4.5 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 47.96% ±1.24 | 44.72% ±1.17 | 3.24 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 8.33% ±2.15 | 12.92% ±2.53 | 4.58 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 30.29% ±3.19 | 37.98% ±3.37 | 7.69 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 77.71% ±0.50 | 85.97% ±0.41 | 8.25 pts | Reported ±1 SE ranges do not overlap |
| EMB | 56.98% ±3.39 | 52.91% ±3.11 | 4.07 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 49.52% ±0.47 | 48.35% ±0.46 | 1.17 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 62.56% ±3.14 | 64.32% ±3.16 | 1.76 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 70.69% ±0.90 | 71.67% ±0.89 | 0.98 pts | Reported ±1 SE ranges overlap |
| MedCode | 41.41% ±2.15 | 43.29% ±2.31 | 1.88 pts | Reported ±1 SE ranges overlap |
| MedScribe | 80.36% ±1.97 | 86.88% ±1.94 | 6.52 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 56.00% ±4.99 | 31.00% ±4.65 | 25.00 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 89.90% ±1.69 | 92.93% ±1.29 | 3.03 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 86.21% ±0.34 | 89.22% ±0.31 | 3.01 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 38.63% ±4.38 | 36.60% ±4.14 | 2.04 pts | Reported ±1 SE ranges overlap |
| IOI | 32.72% ±2.35 | 40.56% ±3.95 | 7.83 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 87.26% ±0.97 | 87.35% ±0.96 | 0.09 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 50.67% ±4.43 | 66.03% ±4.60 | 15.37 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 88.80% ±1.41 | 86.60% ±1.52 | 2.20 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 18.69% ±3.54 | 8.59% ±1.34 | 10.10 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 74.74% ±3.44 | 69.00% ±4.48 | 5.73 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 72.08% ±5.41 | 71.01% ±5.79 | 1.07 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | DeepSeek V4 Flash 0731 average | Grok 4.5 average |
|---|---|---|
| Legal | 38.78% | 45.62% |
| Finance | 59.94% | 59.31% |
| Healthcare | 60.89% | 65.09% |
| Math | 56.00% | 31.00% |
| Academic | 88.05% | 91.07% |
| Coding | 48.94% | 49.34% |
| Cyber | 72.08% | 71.01% |
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 | Grok 4.5 cost | DeepSeek V4 Flash 0731 latency | Grok 4.5 latency |
|---|---|---|---|---|
| Vals Index | $0.85 | $2.46 | 43m06s | 15m44s |
| Harvey's Legal Agent Benchmark | $0.06 | $2.02 | 16m51s | 10m15s |
| Legal Research Bench | $0.28 | $1.09 | 21m07s | 15m54s |
| LegalBench | N/A | N/A | 3.95s | 67.88s |
| EMB | $0.38 | $1.55 | 26m14s | 18m28s |
| Finance Agent (v2) | $0.28 | $1.16 | 8m10s | 4m49s |
| Tax Agent Bench | $0.16 | $0.47 | 5m37s | 4m44s |
| TaxEval v2 | N/A | N/A | 72.20s | 3m15s |
| MedCode | N/A | N/A | 2m07s | 59.64s |
| MedScribe | N/A | N/A | 77.55s | 11m45s |
| ProofBench v1.1 | $0.04 | $0.57 | 15m34s | 11m09s |
| GPQA Diamond | N/A | N/A | 79.96s | 3m38s |
| MMLU Pro | N/A | N/A | 17.19s | 5m58s |
| Code Migration | $3.61 | $3.37 | 2h08m | 23m29s |
| IOI | $0.37 | $5.64 | 27m43s | 1h27m |
| LiveCodeBench | N/A | N/A | 2m07s | 9m49s |
| ProgramBench | $1.29 | $4.12 | 5h39m | 18m59s |
| SkillsBench | $0.13 | $0.58 | 6m52s | 2m16s |
| SWE-bench | $0.01 | $0.54 | 3m22s | 3m20s |
| Terminal-Bench 4.0 | $0.77 | $5.79 | 1h46m | 33m51s |
| Vibe Code Bench v1.1 | $0.91 | $4.10 | 37m22s | 13m50s |
| CyberBench v1.1 | $0.19 | $2.95 | 15m28s | 14m08s |
Results available only for DeepSeek V4 Flash 0731
- BioMysteryBench
- Terminal-Bench Science
Results available only for Grok 4.5
- MortgageTax
- MMMU Pro
- SAGE
- SRE Bench
- Time Horizon Index: KSP