Model comparison
Claude Sonnet 4.6 vs Grok 4.5: Benchmark Comparison
Claude Sonnet 4.6 has the higher score on 10 of 18 shared benchmarks; Grok 4.5 leads on 8.
The largest observed score gap is 21.79 pts on MMMU Pro , where Claude Sonnet 4.6 leads.
Reported ±1 standard-error ranges overlap on 4 of 18 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 | Grok 4.5 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 5.00% ±1.65 | 12.92% ±2.53 | 7.92 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 38.46% ±3.38 | 37.98% ±3.37 | 0.48 pts | Reported ±1 SE ranges overlap |
| LegalBench | 82.12% ±0.48 | 85.97% ±0.41 | 3.84 pts | Reported ±1 SE ranges do not overlap |
| EMB | 60.15% ±3.28 | 52.91% ±3.11 | 7.24 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 51.03% ±0.33 | 48.35% ±0.46 | 2.69 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 67.73% ±0.92 | 61.80% ±0.97 | 5.92 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 64.85% ±3.11 | 64.32% ±3.16 | 0.53 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 77.11% ±0.82 | 71.67% ±0.89 | 5.44 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 85.61% ±2.41 | 92.93% ±1.29 | 7.32 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 87.34% ±0.43 | 89.22% ±0.31 | 1.88 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 83.58% ±0.89 | 61.79% ±1.17 | 21.79 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 46.58% ±3.43 | 34.97% ±3.32 | 11.62 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 39.89% ±4.14 | 36.60% ±4.14 | 3.30 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 82.09% ±1.06 | 87.35% ±0.96 | 5.26 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 0.00% ±0.00 | 0.50 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 49.05% ±4.64 | 66.03% ±4.60 | 16.98 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 77.40% ±1.87 | 86.60% ±1.52 | 9.20 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 51.48% ±4.64 | 69.00% ±4.48 | 17.53 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Sonnet 4.6 average | Grok 4.5 average |
|---|---|---|
| Legal | 41.86% | 45.62% |
| Finance | 64.17% | 59.81% |
| Academic | 85.51% | 81.31% |
| Education | 46.58% | 34.97% |
| Coding | 50.07% | 57.60% |
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 | Grok 4.5 cost | Claude Sonnet 4.6 latency | Grok 4.5 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $3.04 | $2.02 | 18m43s | 10m15s |
| Legal Research Bench | $2.30 | $1.09 | 16m30s | 15m54s |
| LegalBench | N/A | N/A | 21.63s | 67.88s |
| EMB | $7.23 | $1.55 | 54m16s | 18m28s |
| Finance Agent (v2) | $2.41 | $1.16 | 11m36s | 4m49s |
| MortgageTax | N/A | N/A | 29.74s | 117.46s |
| Tax Agent Bench | $1.19 | $0.47 | 11m47s | 4m44s |
| TaxEval v2 | N/A | N/A | 2m08s | 3m15s |
| GPQA Diamond | N/A | N/A | 6m18s | 3m38s |
| MMLU Pro | N/A | N/A | 55.41s | 5m58s |
| MMMU Pro | N/A | N/A | 2m12s | 11m55s |
| SAGE | N/A | N/A | 2m43s | 2m50s |
| Code Migration | $38.51 | $3.37 | 1h40m | 23m29s |
| LiveCodeBench | N/A | N/A | 4m55s | 9m49s |
| ProgramBench | N/A | N/A | 1h12m | 18m59s |
| SkillsBench | $1.38 | $0.58 | 14m20s | 2m16s |
| SWE-bench | $1.30 | $0.54 | 8m32s | 3m20s |
| Vibe Code Bench v1.1 | $5.91 | $4.10 | 26m12s | 13m50s |
Results available only for Claude Sonnet 4.6
- Public Benefits Bench v1.1
Results available only for Grok 4.5
- Vals Index
- MedCode
- MedScribe
- ProofBench v1.1
- IOI
- Terminal-Bench 4.0
- CyberBench v1.1
- SRE Bench
- Time Horizon Index: KSP