DeepSeek V4 Flash 0731 vs GLM 5.2: Benchmark Comparison
DeepSeek V4 Flash 0731 has the higher score on 8 of 16 shared benchmarks; GLM 5.2 leads on 8.
The largest observed score gap is 17.76 pts on LiveCodeBench , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 10 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 | DeepSeek V4 Flash 0731 | GLM 5.2 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 8.33% ±2.15 | 7.08% ±2.00 | 1.25 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 30.29% ±3.19 | 31.25% ±3.22 | 0.96 pts | Reported ±1 SE ranges overlap |
| LegalBench | 77.71% ±0.50 | 84.07% ±0.45 | 6.36 pts | Reported ±1 SE ranges do not overlap |
| EMB | 56.98% ±3.39 | 61.53% ±2.88 | 4.55 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 49.52% ±0.47 | 49.70% ±0.88 | 0.18 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 70.69% ±0.90 | 73.34% ±0.86 | 2.66 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 41.41% ±2.15 | 40.77% ±2.17 | 0.64 pts | Reported ±1 SE ranges overlap |
| MedScribe | 80.36% ±1.97 | 83.53% ±2.00 | 3.17 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 89.90% ±1.69 | 85.61% ±1.77 | 4.29 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 86.21% ±0.34 | 86.71% ±0.33 | 0.51 pts | Reported ±1 SE ranges overlap |
| Code Migration | 38.63% ±4.38 | 37.87% ±4.14 | 0.76 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 87.26% ±0.97 | 69.50% ±1.18 | 17.76 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.00% ±0.00 | 0.50% ±0.50 | 0.50 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 50.67% ±4.43 | 45.08% ±4.39 | 5.59 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 88.80% ±1.41 | 82.80% ±1.69 | 6.00 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 74.74% ±3.44 | 63.96% ±4.79 | 10.78 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4 Flash 0731 average | GLM 5.2 average |
|---|---|---|
| Legal | 38.78% | 40.80% |
| Finance | 59.06% | 61.52% |
| Healthcare | 60.89% | 62.15% |
| Academic | 88.05% | 86.16% |
| Coding | 56.68% | 49.95% |
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 | GLM 5.2 cost | DeepSeek V4 Flash 0731 latency | GLM 5.2 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.06 | $2.06 | 16m51s | 20m51s |
| Legal Research Bench | $0.28 | $0.89 | 21m07s | 17m03s |
| LegalBench | N/A | N/A | 3.95s | 5.77s |
| EMB | $0.38 | $4.47 | 26m14s | 58m52s |
| Finance Agent (v2) | $0.28 | $0.71 | 8m10s | 7m53s |
| TaxEval v2 | N/A | N/A | 72.20s | 64.94s |
| MedCode | N/A | N/A | 2m07s | 95.27s |
| MedScribe | N/A | N/A | 77.55s | 2m18s |
| GPQA Diamond | N/A | N/A | 79.96s | 3m45s |
| MMLU Pro | N/A | N/A | 17.19s | 51.64s |
| Code Migration | $3.61 | $12.64 | 2h08m | 1h54m |
| LiveCodeBench | N/A | N/A | 2m07s | 8m52s |
| ProgramBench | $1.29 | $12.58 | 5h39m | 2h06m |
| SkillsBench | $0.13 | $0.81 | 6m52s | 19m52s |
| SWE-bench | $0.01 | $0.71 | 3m22s | 11m10s |
| Vibe Code Bench v1.1 | $0.91 | $8.49 | 37m22s | 1h02m |
Results available only for DeepSeek V4 Flash 0731
- Vals Index
- Tax Agent Bench
- ProofBench v1.1
- BioMysteryBench
- Terminal-Bench Science
- IOI
- Terminal-Bench 4.0
- CyberBench v1.1
Results available only for GLM 5.2
- SRE Bench