Model comparison
Claude Sonnet 4.6 vs GLM 5.3 Flash: Benchmark Comparison
Claude Sonnet 4.6 has the higher score on 9 of 16 shared benchmarks; GLM 5.3 Flash leads on 7.
The largest observed score gap is 20.72 pts on Vibe Code Bench v1.1 , where Claude Sonnet 4.6 leads.
Reported ±1 standard-error ranges overlap on 9 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 | Claude Sonnet 4.6 | GLM 5.3 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 5.00% ±1.65 | 6.67% ±1.65 | 1.67 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 38.46% ±3.38 | 45.19% ±3.46 | 6.73 pts | Reported ±1 SE ranges overlap |
| LegalBench | 82.12% ±0.48 | 83.93% ±0.43 | 1.81 pts | Reported ±1 SE ranges do not overlap |
| EMB | 60.15% ±3.28 | 55.93% ±3.56 | 4.22 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 51.03% ±0.33 | 57.85% ±2.06 | 6.82 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 64.85% ±3.11 | 61.96% ±3.26 | 2.90 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 77.11% ±0.82 | 75.59% ±0.84 | 1.51 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 85.61% ±2.41 | 86.36% ±2.13 | 0.76 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 87.34% ±0.43 | 86.06% ±0.34 | 1.28 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 83.58% ±0.89 | 86.01% ±0.83 | 2.43 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 39.89% ±4.14 | 20.52% ±4.21 | 19.38 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 82.09% ±1.06 | 80.51% ±1.05 | 1.58 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.50% ±0.50 | 0.00% ±0.00 | 0.50 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 49.05% ±4.64 | 40.16% ±4.43 | 8.89 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 77.40% ±1.87 | 92.00% ±1.21 | 14.60 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 51.48% ±4.64 | 30.76% ±5.28 | 20.72 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Sonnet 4.6 average | GLM 5.3 Flash average |
|---|---|---|
| Legal | 41.86% | 45.26% |
| Finance | 63.29% | 62.83% |
| Academic | 85.51% | 86.14% |
| Coding | 50.07% | 43.99% |
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 | Claude Sonnet 4.6 cost | GLM 5.3 Flash cost | Claude Sonnet 4.6 latency | GLM 5.3 Flash latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $3.04 | $0.57 | 18m43s | 52m17s |
| Legal Research Bench | $2.30 | $0.12 | 16m30s | 51m07s |
| LegalBench | N/A | N/A | 21.63s | 12.52s |
| EMB | $7.23 | $0.71 | 54m16s | 1h15m |
| Finance Agent (v2) | $2.41 | $0.05 | 11m36s | 14m51s |
| Tax Agent Bench | $1.19 | $0.14 | 11m47s | 21m37s |
| TaxEval v2 | N/A | N/A | 2m08s | 53.06s |
| GPQA Diamond | N/A | N/A | 6m18s | 2m11s |
| MMLU Pro | N/A | N/A | 55.41s | 44.76s |
| MMMU Pro | N/A | N/A | 2m12s | 28.22s |
| Code Migration | $38.51 | $3.13 | 1h40m | 3h26m |
| LiveCodeBench | N/A | N/A | 4m55s | 5m02s |
| ProgramBench | N/A | N/A | 1h12m | 5h30m |
| SkillsBench | $1.38 | $0.04 | 14m20s | 12m09s |
| SWE-bench | $1.30 | $0.02 | 8m32s | 1h04m |
| Vibe Code Bench v1.1 | $5.91 | $2.35 | 26m12s | 1h24m |
Results available only for Claude Sonnet 4.6
- MortgageTax
- SAGE
- Public Benefits Bench v1.1
Results available only for GLM 5.3 Flash
- MedScribe
- ProofBench v1.1
- Terminal-Bench Science
- IOI
- Vibe Code Bench 1-100
- CyberBench v1.1