Claude Sonnet 5.5 vs Claude Opus 5.5: Near Opus Level
Anthropic has officially announced Claude Sonnet 5.5 with a significant upgrade over Sonnet 5 and near Opus 5.5 level performance. Here is how Claude Sonnet 5.5 compares against the recently announced Opus 5.5:
Benchmarks
| Benchmark | Claude Sonnet 5.5 | Claude Opus 5.5 | Difference |
|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 66.4% | Sonnet +4.2 points |
| FrontierCode 1.1 (Main) | 52.1%Xhigh effort. 46.2% at Max | 54.4% | Opus +2.3 pointsvs Sonnet at Xhigh |
| CursorBench 4.0 | 55.5% | 57.8% | Opus +2.3 points |
| SWE-Bench Pro | 81.3% | 89.9% | Opus +8.6 points |
| SWE-Bench Multilingual | 90.3% | 93.9% | Opus +3.6 points |
| SWE-Bench Multimodal | 54.3% | 61.4% | Opus +7.1 points |
| GDPval-AA v2.1 | 1844 Elo | 1846 Elo | Opus +2 Elo |
| AA-Briefcase v1.1 | 1811 Elo | 1822 Elo | Opus +11 Elo |
| Humanity's Last Exam, with tools | 64.5% | 67.7% | Opus +3.2 points |
| Humanity's Last Exam, no tools | 56.9% | 64.4% | Opus +7.5 points |
| AutomationBench, Zapier | 44.7% | 42.5% | Sonnet +2.2 points |
| HealthBench Professional | 69.2% | 65.6% | Sonnet +3.6 points |
| Terminal-Bench-Science 0.1 | 59.9% | 58.7% | Sonnet +1.2 points |
| FrontierSWE v2, Proximal | 61.9% | 62.3% | Opus +0.4 points |
| ProgramBench, long context | 79.7% | 91.2% | Opus +11.5 points |
| ArXivMath Aug. 2026, with tools | 95.2% | 96.9% | Opus +1.7 points |
| ArXivMath Aug. 2026, no tools | 86.8% | 91.2% | Opus +4.4 points |
| OSWorld 2.1, partial credit | 80.1% | 81.8% | Opus +1.7 points |
| OSWorld 2.1, strict pass | 43.5% | 48.7% | Opus +5.2 points |
| Chartography, no tools | 61.6% | 64.4% | Opus +2.8 points |
Artificial Analysis benchmarks
| Benchmark | Claude Sonnet 5.5 | Claude Opus 5.5 | Difference |
|---|---|---|---|
| Intelligence Index v4.3.2 | 56 | 58 | Opus +2 |
| AA-Briefcase v1.1 | 1811 Elo | 1822 Elo | Opus +11 Elo |
| GDPval-AA v2.1 | 1844 Elo | 1846 Elo | Opus +2 Elo |
| AutomationBench-AA | 71% | 70% | Sonnet +1 point |
| Terminal-Bench 4.0 | 64% | 60% | Sonnet +4 points |
| SciCode | 61% | 67% | Opus +6 points |
| Humanity's Last Exam | 55% | 61% | Opus +6 points |
| CritPt, physics | 31% | 32% | Opus +1 point |
| AA-Omniscience | 32 | 46 | Opus +14 |
| AA-LCR v1.1, long context | 83% | 85% | Opus +2 points |