Qwen 3.8 Max vs DeepSeek V4 Flash 0731: Benchmark Comparison
Qwen 3.8 Max has the higher score on 16 of 23 shared benchmarks; DeepSeek V4 Flash 0731 leads on 6.
The largest observed score gap is 43.51 pts on CyberBench v1.1 , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 9 of 23 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 | Qwen 3.8 Max | DeepSeek V4 Flash 0731 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 48.27% ±1.34 | 47.96% ±1.24 | 0.31 pts | Reported ±1 SE ranges overlap |
| Harvey's Legal Agent Benchmark | 10.42% ±2.29 | 8.33% ±2.15 | 2.08 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 47.60% ±3.47 | 30.29% ±3.19 | 17.31 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 83.61% ±0.42 | 77.71% ±0.50 | 5.89 pts | Reported ±1 SE ranges do not overlap |
| EMB | 60.07% ±3.35 | 56.98% ±3.39 | 3.09 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 50.59% ±0.58 | 49.52% ±0.47 | 1.08 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 65.96% ±3.18 | 62.56% ±3.14 | 3.40 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 75.55% ±0.84 | 70.69% ±0.90 | 4.87 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 40.67% ±2.03 | 41.41% ±2.15 | 0.75 pts | Reported ±1 SE ranges overlap |
| MedScribe | 84.95% ±2.00 | 80.36% ±1.97 | 4.58 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 58.00% ±4.96 | 56.00% ±4.99 | 2.00 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 12.86% ±4.03 | 4.29% ±2.44 | 8.57 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 93.69% ±1.22 | 89.90% ±1.69 | 3.79 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 88.60% ±0.31 | 86.21% ±0.34 | 2.40 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 23.96% ±4.31 | 38.63% ±4.38 | 14.67 pts | Reported ±1 SE ranges do not overlap |
| IOI | 68.89% ±5.20 | 32.72% ±2.35 | 36.17 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 87.85% ±0.95 | 87.26% ±0.97 | 0.59 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 42.01% ±4.43 | 50.67% ±4.43 | 8.66 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 85.60% ±1.57 | 88.80% ±1.41 | 3.20 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 34.34% ±3.94 | 18.69% ±3.54 | 15.66 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 64.70% ±5.36 | 74.74% ±3.44 | 10.04 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 28.57% ±3.31 | 72.08% ±5.41 | 43.51 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Qwen 3.8 Max average | DeepSeek V4 Flash 0731 average |
|---|---|---|
| Legal | 47.21% | 38.78% |
| Finance | 63.04% | 59.94% |
| Healthcare | 62.81% | 60.89% |
| Math | 58.00% | 56.00% |
| Science | 12.86% | 4.29% |
| Academic | 91.14% | 88.05% |
| Coding | 50.92% | 48.94% |
| Cyber | 28.57% | 72.08% |
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 | Qwen 3.8 Max cost | DeepSeek V4 Flash 0731 cost | Qwen 3.8 Max latency | DeepSeek V4 Flash 0731 latency |
|---|---|---|---|---|
| Vals Index | $4.96 | $0.85 | 1h41m | 43m06s |
| Harvey's Legal Agent Benchmark | $2.37 | $0.06 | 49m12s | 16m51s |
| Legal Research Bench | $2.49 | $0.28 | 1h01m | 21m07s |
| LegalBench | N/A | N/A | 36.37s | 3.95s |
| EMB | $2.66 | $0.38 | 57m47s | 26m14s |
| Finance Agent (v2) | $1.24 | $0.28 | 19m35s | 8m10s |
| Tax Agent Bench | $1.55 | $0.16 | 40m36s | 5m37s |
| TaxEval v2 | N/A | N/A | 4m26s | 72.20s |
| MedCode | N/A | N/A | 6m18s | 2m07s |
| MedScribe | N/A | N/A | 4m11s | 77.55s |
| ProofBench v1.1 | $1.50 | $0.04 | 31m08s | 15m34s |
| Terminal-Bench Science | $16.22 | $1.10 | 3h20m | 1h34m |
| GPQA Diamond | N/A | N/A | 4m27s | 79.96s |
| MMLU Pro | N/A | N/A | 63.58s | 17.19s |
| Code Migration | $10.54 | $3.61 | 4h28m | 2h08m |
| IOI | $9.47 | $0.37 | 1h47m | 27m43s |
| LiveCodeBench | N/A | N/A | 5m38s | 2m07s |
| ProgramBench | $12.36 | $1.29 | 5h54m | 5h39m |
| SkillsBench | $0.52 | $0.13 | 15m48s | 6m52s |
| SWE-bench | $1.13 | $0.01 | 41m42s | 3m22s |
| Terminal-Bench 4.0 | $10.63 | $0.77 | 2h28m | 1h46m |
| Vibe Code Bench v1.1 | $8.24 | $0.91 | 2h49m | 37m22s |
| CyberBench v1.1 | $0.78 | $0.19 | 9m40s | 15m28s |
Results available only for Qwen 3.8 Max
- Vals RSI Index
- MortgageTax
- MysteryMechanism
- MMMU Pro
- SAGE
- Vibe Code Bench 1-100
- Public Benefits Bench v1.1
Results available only for DeepSeek V4 Flash 0731
- BioMysteryBench