Claude Opus 4.8 vs Claude Opus 5
Vals Science Index?
Opus 4.8—
Opus 547.9
| Benchmark | Opus 4.8 | Opus 5 |
|---|---|---|
| Humanity's Last Exam | 48.7 | 51.3 |
| Arena Text · overall | 1,481 · high | 1,490 · max |
| Arena Text · Chinese | 1,506.8 | 1,573.4 · max |
| Arena Text · English | 1,468.6 · high | 1,507.3 · max |
| Arena Text · French | 1,481.1 · high | 1,519.5 · high |
| Arena Text · German | 1,472.4 · high | 1,525.9 · max |
| Arena Text · instruction following | 1,476.3 · high | 1,517.8 · max |
| Arena Text · Japanese | 1,441.1 · high | 1,515.4 · max |
| Arena Text · Korean | 1,431.9 · high | 1,520.7 · max |
| Arena Text · non-English | 1,450 · high | 1,500.5 · max |
| Arena Text · Polish | 1,459.3 | 1,519.9 · max |
| Arena Text · Russian | 1,474 · high | 1,507.2 · max |
| Arena Text · Spanish | 1,465 · high | 1,519.5 · max |
| AA Omniscience Index | 28.8 | 31 |
| SimpleQA Verified v1 | 53 | 63.3 |
| CritPt (physics) | 20.9 | 26.9 |
| AA Intelligence Index 4.3.2 | 41.8 | 44.8 |
| AA Omniscience · Accuracy | 48.8 | 57.1 |
| AA Omniscience · Hallucination Rate | 39.3 | 60.7 |
| AA-LCR (long context) 1.1 | 77.7 | 82 |
| GDP.pdf · AA all-pass | 22.8 | 20 |
| MLCR-AA | 45.6 | 56.1 |
| ARC-AGI-1 · semi-private | 92.5 | 97.5 · max |
| ARC-AGI-2 · semi-private | 72.1 · high | 90.4 · max |
| ARC-AGI-3 · semi-private | 1.5 · high | 30.2 · high |
| Chess Puzzles | 52 | 43 · max |
| FrontierMath Tier 4 (v2) | 56.1 | 73.2 · max |
| FrontierMath Tiers 1–3 (v2) | 80 | 85.6 · max |
| Mystery Game Puzzles | 36 | 59 · max |
| MathArena · ArXivMath · June 2026 | 70 | 82.6 · max |
| MathArena · BrokenArXiv · June 2026 | 41.7 | 90.7 · max |
| MathArena · ArXivLean · June 2026 | 23.9 | 32.6 · max |
| Vals LegalBench | 83.6 | 87 |
| Vals MedCode | 53.2 | 63.6 |
| Vals MedScribe | 85.8 | 91 |
| Vals Legal Research Bench | 43.8 | 55.3 |
| IFBench (instructions) | 62.2 | — |
| ARC-AGI-1 · public evaluation | — | 99 · max |
| ARC-AGI-2 · public evaluation | — | 97.1 · max |
| EQ-Bench 3 | 2,029.8 | — |
| FrontierMath | 47.2 | — |
| FrontierMath Erdos | — | 0 · max |
| FrontierMath Tier 4 | 31.3 | — |
| MathArena · Apex | 81.3 | — |
| MathArena · Apex Shortlist | 90.4 | — |
| MathArena · ArXivMath · April 2026 | 62.5 | — |
| MathArena · ArXivMath · February 2026 | 60.2 | — |
| MathArena · ArXivMath · March 2026 | 75 | — |
| MathArena · ArXivMath · May 2026 | 65 | — |
| MathArena · BrokenArXiv · April 2026 | 34.4 | — |
| MathArena · BrokenArXiv · February 2026 | 35.5 | — |
| MathArena · BrokenArXiv · March 2026 | 35.7 | — |
| MathArena · BrokenArXiv · May 2026 | 23 | — |
| MathArena · Kangaroo 2025 1-2 | 70.8 | — |
| MathArena · Kangaroo 2025 11-12 | 100 | — |
| MathArena · Kangaroo 2025 3-4 | 60.4 | — |
| MathArena · Kangaroo 2025 5-6 | 70 | — |
| MathArena · Kangaroo 2025 7-8 | 96.7 | — |
| MathArena · Kangaroo 2025 9-10 | 91.7 | — |
| Vals Science Index | — | 47.9 |
Specs
| Spec | Opus 4.8 | Opus 5 |
|---|---|---|
| Input, per 1M | $5 | $5 |
| Output, per 1M | $25 | $25 |
| Cost per task? | $4.08 | $2.19 |
| Speed, tok/s | 66 | 55 |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | May 2026 | Jul 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | Partial |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
Opus 4.8 → Opus 5
Nothing known.
Opus 5 → Opus 4.8
Nothing known.