Model comparison
Claude Sonnet 5.5 vs GPT-5.6 Terra: Benchmark Comparison
Claude Sonnet 5.5 has the higher score on 15 of 17 shared benchmarks; GPT-5.6 Terra leads on 2.
The largest observed score gap is 28.57 pts on Terminal-Bench Science , where Claude Sonnet 5.5 leads.
Reported ±1 standard-error ranges overlap on 4 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 | Claude Sonnet 5.5 | GPT-5.6 Terra | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 69.22% ±0.96 | 59.59% ±1.33 | 9.63 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 2.92% ±1.36 | 0.83% ±0.83 | 2.08 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 48.08% ±3.47 | 41.35% ±3.42 | 6.73 pts | Reported ±1 SE ranges overlap |
| EMB | 75.71% ±2.44 | 66.20% ±3.11 | 9.51 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 58.10% ±0.67 | 54.44% ±2.07 | 3.67 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 73.39% ±2.98 | 65.20% ±3.19 | 8.20 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 52.92% ±2.12 | 43.41% ±2.17 | 9.51 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 91.10% ±1.96 | 82.87% ±1.95 | 8.23 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 100.00% ±0.00 | 74.00% ±4.41 | 26.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 38.57% ±5.86 | 10.00% ±3.61 | 28.57 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 51.77% ±3.41 | 47.00% ±3.40 | 4.77 pts | Reported ±1 SE ranges overlap |
| Code Migration | 69.83% ±4.26 | 47.80% ±4.28 | 22.02 pts | Reported ±1 SE ranges do not overlap |
| IOI | 83.06% ±3.74 | 87.61% ±6.53 | 4.56 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 53.03% ±1.51 | 26.26% ±0.51 | 26.77 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 92.39% ±1.26 | 74.59% ±4.20 | 17.80 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 59.58% ±5.21 | 72.08% ±5.41 | 12.50 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 67.19% ±1.22 | 62.38% ±1.26 | 4.80 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Sonnet 5.5 average | GPT-5.6 Terra average |
|---|---|---|
| Index | 69.22% | 59.59% |
| Legal | 25.50% | 21.09% |
| Finance | 69.07% | 61.95% |
| Healthcare | 72.01% | 63.14% |
| Math | 100.00% | 74.00% |
| Science | 38.57% | 10.00% |
| Education | 51.77% | 47.00% |
| Coding | 74.58% | 59.07% |
| Beta | 59.58% | 72.08% |
| Social Mobility | 67.19% | 62.38% |
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 5.5 cost | GPT-5.6 Terra cost | Claude Sonnet 5.5 latency | GPT-5.6 Terra latency |
|---|---|---|---|---|
| Vals Index | $20.80 | $5.18 | 1h10m | 28m50s |
| Harvey's Legal Agent Benchmark | $16.61 | $3.20 | 57m44s | 17m05s |
| Legal Research Bench | $14.83 | $7.91 | 1h33m | 1h10m |
| EMB | $8.79 | $2.19 | 36m45s | 13m33s |
| Finance Agent (v2) | $6.55 | $3.63 | 31m43s | 24m10s |
| Tax Agent Bench | $9.25 | $4.83 | 59m21s | 1h01m |
| MedCode | N/A | N/A | 3m59s | 18.41s |
| MedScribe | N/A | N/A | 5m07s | 35.53s |
| ProofBench v1.1 | $0.56 | $1.27 | 5m48s | 12m57s |
| Terminal-Bench Science | $31.98 | $10.58 | 3h22m | 2h47m |
| SAGE | N/A | N/A | 113.42s | 40.06s |
| Code Migration | $75.83 | $8.13 | 3h20m | 48m34s |
| IOI | $7.52 | $8.66 | 45m51s | 1h23m |
| Terminal-Bench 4.0 | $19.33 | $5.91 | 2h06m | 1h27m |
| Vibe Code Bench v1.1 | $31.25 | $7.89 | 1h01m | 19m12s |
| CyberBench v1.1 | $4.42 | $3.31 | 33m53s | 17m04s |
| Public Benefits Bench v1.1 | $4.25 | $1.20 | 1h18m | 21m54s |
Results available only for Claude Sonnet 5.5
- BioMysteryBench
- MysteryMechanism
- SRE Bench
Results available only for GPT-5.6 Terra
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- ProgramBench
- SkillsBench
- SWE-bench
- Vibe Code Bench 1-100