Claude Opus 5 vs GPT-6.2 Astra
GPT-6.2 Astra leads by 4.6 capability points (≈95% interval +2.0 to +7.2): a clear lead over Claude Opus 5 · P(higher) ≈ 100%
Wins 46 of 55 shared benchmarks (9 lost)
Capability vs. relative cost
Claude Opus 5GPT-6.2 AstraClosed weightsOpen weights
Benchmark wins9 – 46 · 55 shared
Comparing matching setups only (15 benchmarks excluded) · suites counted once
Area← Claude Opus 5GPT-6.2 Astra →
Data: Artificial Analysis, Epoch AI, Vals AI and 12 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Astra minus Claude Opus 5.
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 28.6 | 78.1 | +49.5 |
| EBR-bench v4 (card ban, single agent) | Games | 45.7 | 92.4 | +46.7 |
| Terminal-Bench Science 0.1 (Vals) | Science › Multidisciplinary | 27.1 | 71.4 | +44.3 |
| Mystery Game Puzzles | Games | 59.0 | 95.0 | +36.0 |
| Chess Puzzles (task version 2.0) | Games | 43.0 | 78.0 | +35.0 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 30.5 | 63.2 | +32.7 |
| Blueprint-Bench 2 | Perception & media › Spatial & visual reasoning | 0.3 | 0.6 | +31.0 |
| DeepsecBench — Revalidated V1 (233 issues) | Software › Cybersecurity | 32.3 | 61.2 | +28.9 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 60.8 | 89.2 | +28.3 |
| ARC-AGI-3 (Semi-Private) | Games | 30.2 | 58.0 | +27.8 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 44.8 | 72.2 | +27.4 |
| FrontierMath Tier 4 (v2) - Private | Math | 73.2 | 100.0 | +26.8 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 13.9 | 39.4 | +25.6 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 26.9 | 52.4 | +25.4 |
| EnterpriseBench: CoreCraft | Tool & computer use › Customer service & operations | 68.7 | 43.6 | −25.1 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 20.0 | 44.0 | +24.0 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 23.8 | 45.8 | +22.0 |
| DeepsecBench-AA (Artificial Analysis) | Software › Cybersecurity | 25.9 | 46.8 | +20.9 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 56.6 | 77.2 | +20.7 |
| Vals Science Index | Science › Multidisciplinary | 47.9 | 68.4 | +20.5 |
| MirrorCode (ML, +Private, 2L) | Software › Software engineering | 87.8 | 67.8 | −20.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 76.7 | 57.9 | −18.8 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 44.2 | 62.1 | +18.0 |
| EnterpriseOps-Gym-AA | Tool & computer use › Customer service & operations | 47.5 | 29.9 | −17.6 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 6.3 | 15.2 | +17.5 |
| MysteryMechanism | Math | 37.4 | 53.2 | +15.8 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 49.0 | 64.1 | +15.2 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 31.2 | 45.4 | +14.2 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 10.3 | 37.0 | +13.4 |
| Legal Research Bench | Professional › Legal | 55.3 | 42.8 | −12.5 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 22.0 | 34.4 | +12.4 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 71.4 | 83.7 | +12.3 |
| FrontierMath Erdős | Math | 0.0 | 10.3 | +10.3 |
| GPQA Diamond | Science › Multidisciplinary | 87.9 | 96.8 | +9.0 |
| AA-AnalystAgent (pass^5) | Professional › General knowledge work | 53.8 | 62.5 | +8.8 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 21.6 | 29.8 | +8.2 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 85.6 | 92.6 | +7.0 |
| Epoch Capabilities Index (ECI) | Composite indexes | 165 | 173 | +6.5 |
| Chartography (November 2026 rescoring) | Perception & media › Documents & charts | 84.4 | 90.7 | +6.3 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 37.6 | 50.0 | +6.2 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 2.4 | 14.0 | +5.8 |
| MMMU Pro | Knowledge & reasoning › General knowledge | 84.0 | 89.5 | +5.5 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 90.4 | 95.8 | +5.4 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 63.0 | 68.0 | +5.0 |
| BigFinanceBench private holdout (final-answer accuracy) | Professional › Finance, accounting & tax | 49.2 | 54.0 | +4.8 |
| LLM Debate Benchmark | Language & writing › Language & wordplay | 1704 | 1673 | −4.4 |
| SimpleBench | Knowledge & reasoning › General knowledge | 80.6 | 77.3 | −3.3 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 88.4 | 89.9 | +1.5 |
| BioMysteryBench (Vals Terminus 2) | Science › Life sciences | 79.3 | 80.7 | +1.5 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 10.1 | 12.4 | +1.2 |
| OTIS Mock AIME 2024-2025 | Math | 98.9 | 100.0 | +1.1 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 97.5 | 98.5 | +1.0 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1654 | 1644 | −0.6 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 8.7 | 7.7 | −0.5 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.0 | 1.2 | +0.1 |
Data: Artificial Analysis, Epoch AI, Vals AI and 12 othersatlas.village.dev