Claude Opus 4.8 vs Claude Opus 5
Claude Opus 4.8 Verdict
- Claude Opus 5 leads on quality: 70.0 vs 64.7.
- Claude Opus 4.8 is stronger in instruction following.
- Claude Opus 5 is stronger in agents & tools, coding, composite, human preference, knowledge, long context, multimodal, reasoning, maths.
- They cost about the same ($10.00 per 1M blended).
| Metric | Claude Opus 4.8 | Claude Opus 5 |
|---|---|---|
| BenchLeader Index | 64.7 | 70.0 |
| Agents & tools score | 65.0 | 71.6 |
| Coding score | 65.7 | 74.4 |
| Composite score | 81.9 | 92.9 |
| Human preference score | 65.6 | 78.2 |
| Instruction following score | 61.6 | – |
| Knowledge score | 66.3 | 70.5 |
| Long context score | 65.3 | 66.2 |
| Multimodal score | 64.4 | 69.4 |
| Reasoning score | 64.2 | 71.1 |
| Maths score | – | 67.1 |
| Blended price $/M | $10.00 | $10.00 |
| Output speed | 58 tok/s | 52 tok/s |
| Time to first answer | 29.5 s | 81.6 s |
| Context window | 1M | 1M |
| GPQA Diamond | – | 92.9% |
| OTIS Mock AIME | – | 97.8% |
| SimpleBench | 64.8% | 80.6% |
| Remote Labor Index | 8.3% | – |
| WeirdML | – | 86.3% |
| FrontierCode | 46.5% | – |
| GSO-Bench | 47.1% | – |
| Epoch Capabilities Index | 158.3 | 162.6 |
| LMArena Text | 1473 | – |
| LMArena Hard Prompts | 1503 | – |
| LMArena Coding | 1527 | – |
| LMArena WebDev | 1540 | – |
| LMArena Vision | 1289 | – |
| AA Intelligence Index | 42.0 | 50.7 |
| IFBench | 62.2% | – |
| AA-LCR | 77.7% | 79.3% |
| MMMU-Pro | – | 84.7% |
| AA-Omniscience | 28.8 | 37.1 |
| Terminal-Bench Hard | 58.3% | – |
| GPQA Diamond (AA) | 92.0% | 93.2% |
| Humanity's Last Exam (AA) | 48.7% | 54.9% |
| SciCode (AA) | 54.4% | 56.4% |
| τ²-Bench Telecom (AA) | 94.4% | – |
| LiveCodeBench | 87.8% | 89.0% |
| MMLU-Pro | 89.6% | 91.6% |
| IOI | – | 91.7% |
| LegalBench | 83.6% | 87.0% |
| CorpFin | 66.7% | 73.2% |
| TaxEval | 75.6% | 75.1% |
| Terminal-Bench 2.1 (Vals) | 71.9% | 84.6% |
| SWE-bench (Vals) | 88.6% | 97.0% |
| GPQA Diamond (Vals) | 92.4% | 93.4% |
| Vals Index | 60.9 | 67.2 |
| HiL-Bench | 35.3% | 57.0% |
| EQ-Bench 4 | 1281 | 1385 |
Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.