DeepSeek V4 Flash 0731 vs GLM 5.3 Flash: Benchmark Comparison
DeepSeek V4 Flash 0731 has the higher score on 8 of 17 shared benchmarks; GLM 5.3 Flash leads on 8.
The largest observed score gap is 43.98 pts on Vibe Code Bench v1.1 , where DeepSeek V4 Flash 0731 leads.
Reported ±1 standard-error ranges overlap on 6 of 17 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 5.3 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 30.29% ±3.19 | 45.19% ±3.46 | 14.90 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 77.71% ±0.50 | 83.93% ±0.43 | 6.21 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 28.07% ±3.16 | 31.55% ±3.35 | 3.48 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 70.69% ±0.90 | 75.59% ±0.84 | 4.91 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 80.36% ±1.97 | 88.94% ±1.91 | 8.57 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 56.00% ±4.99 | 21.00% ±4.09 | 35.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench-Science | 4.29% ±2.44 | 2.86% ±2.01 | 1.43 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 89.90% ±1.69 | 86.36% ±2.13 | 3.53 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 86.21% ±0.34 | 86.06% ±0.34 | 0.15 pts | Reported ±1 SE ranges overlap |
| IOI | 32.72% ±2.35 | 52.50% ±2.62 | 19.78 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 87.26% ±0.97 | 80.51% ±1.05 | 6.75 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 |
| SkillsBench | 50.67% ±4.43 | 40.16% ±4.43 | 10.51 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 88.80% ±1.41 | 92.00% ±1.21 | 3.20 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 18.69% ±3.54 | 37.37% ±2.67 | 18.69 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 74.74% ±3.44 | 30.76% ±5.28 | 43.98 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 72.08% ±5.41 | 67.86% ±5.57 | 4.23 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | DeepSeek V4 Flash 0731 average | GLM 5.3 Flash average |
|---|---|---|
| Legal | 54.00% | 64.56% |
| Finance | 49.38% | 53.57% |
| Healthcare | 80.36% | 88.94% |
| Math | 56.00% | 21.00% |
| Science | 4.29% | 2.86% |
| Academic | 88.05% | 86.21% |
| Coding | 50.41% | 47.61% |
| Cyber | 72.08% | 67.86% |
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 5.3 Flash cost | DeepSeek V4 Flash 0731 latency | GLM 5.3 Flash latency |
|---|---|---|---|---|
| Legal Research Bench | $0.28 | $0.12 | 21m07s | 51m07s |
| LegalBench | N/A | N/A | 3.95s | 12.52s |
| Tax Agent Bench | $0.16 | $0.14 | 5m37s | 21m37s |
| TaxEval v2 | N/A | N/A | 72.20s | 53.06s |
| MedScribe | N/A | N/A | 77.55s | 86.98s |
| ProofBench v1.1 | $0.04 | $0.22 | 15m34s | 1h11m |
| Terminal-Bench-Science | $1.10 | $0.52 | 1h34m | 3h28m |
| GPQA Diamond | N/A | N/A | 79.96s | 2m11s |
| MMLU Pro | N/A | N/A | 17.19s | 44.76s |
| IOI | $0.37 | $0.43 | 27m43s | 1h32m |
| LiveCodeBench | N/A | N/A | 2m07s | 5m02s |
| ProgramBench | $1.29 | $0.80 | 5h39m | 5h30m |
| SkillsBench | $0.13 | $0.04 | 6m52s | 12m09s |
| SWE-bench | $0.01 | $0.02 | 3m22s | 1h04m |
| Terminal-Bench 4.0 | $0.77 | $1.22 | 1h46m | 1h01m |
| Vibe Code Bench v1.1 | $0.91 | $2.35 | 37m22s | 1h24m |
| CyberBench v1.1 | $0.19 | $0.14 | 15m28s | 35m48s |
Results available only for DeepSeek V4 Flash 0731
- MedCode
- BioMysteryBench
Results available only for GLM 5.3 Flash
- Vals Finance Index
- Vals Index
- Vals Legal Index
- HLAB-Vals-1.1
- EMB v1.1
- Finance Agent v2.1
- MMMU Pro
- Code Migration v1.1
- Vibe Code Bench 1-100
- SRE Bench