GPT-5.6 Terra vs GPT-6.2 Astra
GPT-6.2 Astra leads by 8.1 capability points (≈95% interval +5.7 to +10.6): a clear lead over GPT-5.6 Terra · P(higher) ≈ 100%
Wins 52 of 54 shared benchmarks (2 lost)
Capability vs. relative cost
GPT-5.6 TerraGPT-6.2 AstraClosed weightsOpen weights
Benchmark wins2 – 52 · 54 shared
Comparing matching setups only (4 benchmarks excluded) · suites counted once
Area← GPT-5.6 TerraGPT-6.2 Astra →
Data: Artificial Analysis, Epoch AI, Arena and 11 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Astra minus GPT-5.6 Terra.
| ARC-AGI-3 (Semi-Private) | Games | 0.8 | 70.3 | +69.5 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 13.8 | 78.1 | +64.3 |
| Terminal-Bench Science 0.1 (Vals) | Science › Multidisciplinary | 10.0 | 71.4 | +61.4 |
| Mystery Game Puzzles | Games | 35.0 | 95.0 | +60.0 |
| ExploitBench - release-reported exploit progress | Software › Cybersecurity | 46.2 | 100.0 | +53.8 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 28.8 | 72.2 | +43.4 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 3.3 | 45.8 | +42.5 |
| DeepsecBench — Revalidated V1 (233 issues) | Software › Cybersecurity | 19.1 | 61.2 | +42.2 |
| MedChemBench | Science › Life sciences | 43.3 | 84.5 | +41.2 |
| Arena Code WebDev — HTML | Software › Frontend & design | 1523 | 1827 | +40.7 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 22.7 | 63.2 | +40.5 |
| Terminal-Bench 4.0 | Software › Software engineering | 22.7 | 63.1 | +40.4 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 14.8 | 52.4 | +37.6 |
| Arena Code WebDev — React | Software › Frontend & design | 1511 | 1784 | +37.5 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 54.2 | 89.2 | +35.0 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 37.3 | 2.4 | +34.9 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | -2.7 | 15.2 | +34.9 |
| Chartography (November 2026 rescoring) | Perception & media › Documents & charts | 58.8 | 90.7 | +31.9 |
| Chess Puzzles (task version 2.0) | Games | 47.0 | 78.0 | +31.0 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 3.6 | 34.4 | +30.8 |
| Blueprint-Bench 2 | Perception & media › Spatial & visual reasoning | 0.3 | 0.6 | +30.6 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 32.3 | 62.1 | +29.8 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | -9.3 | 50.0 | +29.6 |
| FrontierMath Tier 4 (v2) - Private | Math | 70.7 | 100.0 | +29.3 |
| Bug Hunt Bench | Software › Software engineering | 30.5 | 59.7 | +29.2 |
| SEC-Bench Pro - release-reported cohort | Software › Cybersecurity | 57.7 | 86.8 | +29.0 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 35.4 | 64.1 | +28.8 |
| MirrorCode (ML, +Private, 2L) | Software › Software engineering | 41.1 | 67.8 | +26.7 |
| DeepsecBench-AA (Artificial Analysis) | Software › Cybersecurity | 25.1 | 46.8 | +21.7 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 23.9 | 45.4 | +21.5 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | -3.3 | 37.0 | +20.2 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 67.2 | 84.9 | +17.7 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 59.6 | 77.2 | +17.6 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1333 | 1644 | +17.0 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 74.6 | 89.9 | +15.3 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 69.7 | 83.7 | +14.0 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 82.5 | 95.8 | +13.3 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 27.2 | 39.4 | +12.2 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | -7.7 | 12.4 | +10.1 |
| MMMU Pro | Knowledge & reasoning › General knowledge | 79.5 | 89.5 | +10.0 |
| Epoch Capabilities Index (ECI) | Composite indexes | 161 | 173 | +9.7 |
| EnterpriseOps-Gym-AA | Tool & computer use › Customer service & operations | 38.5 | 29.9 | −8.6 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | -2.0 | 14.0 | +8.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 65.2 | 57.9 | −7.3 |
| MentalHealthBench | Health & medicine | 46.2 | 53.3 | +7.1 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 86.0 | 92.6 | +6.7 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 24.6 | 29.8 | +5.2 |
| GPQA Diamond | Science › Multidisciplinary | 92.9 | 97.0 | +4.1 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -0.3 | 7.7 | +4.0 |
| HealthBench Professional | Health & medicine | 62.4 | 66.0 | +3.6 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 96.5 | 98.5 | +2.0 |
| Legal Research Bench | Professional › Legal | 41.3 | 42.8 | +1.4 |
| OTIS Mock AIME 2024-2025 | Math | 99.7 | 100.0 | +0.3 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.2 | 1.2 | 0.0 |
Data: Artificial Analysis, Epoch AI, Arena and 11 othersatlas.village.dev