Gemini 4 Hydrogen vs GPT-6.2 Astra
GPT-6.2 Astra leads by 2.8 capability points (≈95% interval +0.1 to +5.6): a clear lead over Gemini 4 Hydrogen · P(higher) ≈ 98%
Wins 52 of 74 shared benchmarks (19 lost, 3 tied)
Capability vs. relative cost
Gemini 4 HydrogenGPT-6.2 AstraClosed weightsOpen weights
Benchmark wins19 – 52 · 74 shared, 3 tied
Comparing matching setups only (31 benchmarks excluded) · suites counted once
Area← Gemini 4 HydrogenGPT-6.2 Astra →
Data: Artificial Analysis, Epoch AI, Surge AI and 17 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Astra minus Gemini 4 Hydrogen.
| KradleBench | Games | 43.9 | 78.3 | +34.4 |
| SpireClimb | Games | 7.7 | 41.3 | +33.7 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 9.7 | 41.4 | +31.7 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 57.5 | 89.2 | +31.7 |
| Mystery Game Puzzles | Games | 65.0 | 95.0 | +30.0 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 49.5 | 78.6 | +29.0 |
| EBR-bench v4 (card ban, single agent) | Games | 70.5 | 92.4 | +21.9 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 41.2 | 62.1 | +20.9 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 78.4 | 57.9 | −20.6 |
| Chess Puzzles (task version 2.0) | Games | 60.0 | 78.0 | +18.0 |
| SimpleBench | Knowledge & reasoning › General knowledge | 91.7 | 77.3 | −14.4 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 58.8 | 72.2 | +13.4 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 27.2 | 39.4 | +12.2 |
| FrontierMath Tier 4 (v2) - Private | Math | 87.8 | 100.0 | +12.2 |
| GDP.xlsx | Professional › General knowledge work | 40.9 | 29.7 | −11.2 |
| Legal Research Bench | Professional › Legal | 53.8 | 42.8 | −11.1 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 53.5 | 64.1 | +10.6 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 24.0 | 34.4 | +10.4 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 53.1 | 43.1 | −10.0 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 39.0 | 49.0 | +10.0 |
| Terminal-Bench 4.0 | Software › Software engineering | 54.0 | 63.1 | +9.1 |
| Vals Finance Index | Professional › Finance, accounting & tax | 66.1 | 57.3 | −8.8 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 33.6 | 42.0 | +8.4 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 77.0 | 84.9 | +7.9 |
| SimpleQA Verified (Epoch anti-abstention) | Knowledge & reasoning › General knowledge | 70.9 | 78.5 | +7.6 |
| MultiChallenge | Language & writing › Conversation | 80.6 | 73.2 | −7.4 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 38.5 | 45.8 | +7.3 |
| Agents' Last Exam v1.1 Score | Professional › General knowledge work | 71.6 | 78.3 | +6.7 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 61.3 | 68.0 | +6.7 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 31.7 | 38.3 | +6.6 |
| Artificial Analysis Engineering Index | Science › Engineering | 60.1 | 66.6 | +6.5 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -4.5 | 7.7 | +6.1 |
| Artificial Analysis Economics Index | Composite indexes | 55.0 | 61.1 | +6.0 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 3.6 | 9.2 | +5.6 |
| Artificial Analysis Strategy & Ops Index | Professional › General knowledge work | 57.7 | 63.3 | +5.5 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 83.2 | 78.2 | −5.0 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 54.2 | 49.2 | −5.0 |
| HLAB-Vals-1.1 | Professional › Legal | 11.3 | 16.3 | +5.0 |
| Epoch Capabilities Index (ECI) | Composite indexes | 167 | 173 | +4.8 |
| BigFinanceBench private holdout (final-answer accuracy) | Professional › Finance, accounting & tax | 49.4 | 54.0 | +4.6 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 88.1 | 92.6 | +4.6 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 65.1 | 60.7 | −4.4 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 70.8 | 75.0 | +4.2 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 73.7 | 69.6 | −4.1 |
| Vals Legal Index | Professional › Legal | 73.6 | 69.7 | −3.9 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 29.3 | 37.0 | +3.9 |
| Artificial Analysis Healthcare & Medical Index | Health & medicine | 53.3 | 57.0 | +3.7 |
| DrugDiscoveryBench | Science › Life sciences | 69.1 | 65.5 | −3.6 |
| Chartography (November 2026 rescoring) | Perception & media › Documents & charts | 87.2 | 90.7 | +3.5 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 92.5 | 95.8 | +3.3 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 13.5 | 15.2 | +3.3 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 32.8 | 29.8 | −3.0 |
| Artificial Analysis Finance & Accounting Index | Professional › Finance, accounting & tax | 58.1 | 60.9 | +2.9 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 74.4 | 77.2 | +2.8 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 80.6 | 83.1 | +2.6 |
| Dual-Arm Sim (task completion) | Tool & computer use › Robotics | 0.0 | 2.5 | +2.5 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 0.0 | 2.4 | −2.4 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 87.8 | 89.9 | +2.1 |
| Artificial Analysis Legal Index | Professional › Legal | 56.7 | 58.7 | +2.0 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 46.3 | 50.0 | +1.9 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 17.4 | 14.0 | −1.7 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 15.7 | 12.4 | −1.6 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1569 | 1594 | +1.4 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 51.1 | 52.4 | +1.2 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 44.6 | 45.8 | +1.2 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 82.7 | 83.6 | +0.9 |
| Vals Index v2.2 | Composite indexes | 68.5 | 67.7 | −0.8 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.7 | 1.2 | +0.2 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 75.8 | 76.0 | +0.2 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 63.4 | 63.2 | −0.1 |
| GPQA Diamond | Science › Multidisciplinary | 96.8 | 96.8 | +0.1 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 98.5 | 98.5 | 0.0 |
| OTIS Mock AIME 2024-2025 | Math | 100.0 | 100.0 | 0.0 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 44.0 | 44.0 | 0.0 |
Data: Artificial Analysis, Epoch AI, Surge AI and 17 othersatlas.village.dev