Claude Opus 5 vs Gemini 4 Hydrogen
TryTop 2 now
Gemini 4 Hydrogen leads by 1.8 capability points (≈95% interval −0.7 to +4.3): statistically tied · P(higher) ≈ 92%
Wins 43 of 55 shared benchmarks (12 lost)
Capability vs. relative cost
Claude Opus 5Gemini 4 HydrogenClosed weightsOpen weights
Benchmark wins12 – 43 · 55 shared
Comparing matching setups only (3 benchmarks excluded) · suites counted once
Area← Claude Opus 5Gemini 4 Hydrogen →
Data: Artificial Analysis, Epoch AI, Arena and 19 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Hydrogen minus Claude Opus 5.
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 30.5 | 63.4 | +32.9 |
| EBR-bench v4 (card ban, single agent) | Games | 45.7 | 70.5 | +24.8 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 26.9 | 51.1 | +24.2 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 20.0 | 44.0 | +24.0 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 28.6 | 51.4 | +22.8 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 56.6 | 74.4 | +17.9 |
| Chess Puzzles (task version 2.0) | Games | 43.0 | 60.0 | +17.0 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 23.8 | 38.5 | +14.7 |
| FrontierMath Tier 4 (v2) - Private | Math | 73.2 | 87.8 | +14.6 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 6.3 | 13.5 | +14.1 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 44.8 | 58.8 | +14.0 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 13.9 | 27.2 | +13.3 |
| FrontierCode 2.0 - Pass Rate | Software › Software engineering | 31.4 | 43.2 | +11.8 |
| TxBench Oligo Discovery V1 (113 evaluations) | Science › Life sciences | 42.5 | 54.3 | +11.8 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 21.6 | 32.8 | +11.2 |
| SimpleBench | Knowledge & reasoning › General knowledge | 80.6 | 91.7 | +11.1 |
| FrontierCode 2.0 - Score | Software › Software engineering | 28.1 | 38.9 | +10.7 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 10.3 | 29.3 | +9.5 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 71.4 | 80.6 | +9.1 |
| TxBench Antibody Discovery V1 (100 evaluations) | Science › Life sciences | 48.7 | 57.7 | +9.0 |
| GPQA Diamond | Science › Multidisciplinary | 87.9 | 96.8 | +8.9 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 46.0 | 54.0 | +8.1 |
| CursorBench 4.0 | Software › Software engineering | 46.6 | 54.3 | +7.7 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 2.4 | 17.4 | +7.5 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 8.7 | -4.5 | −6.6 |
| Mystery Game Puzzles | Games | 59.0 | 65.0 | +6.0 |
| BuseyBench SVG | Perception & media › Generation | 6.3 | 6.9 | +5.6 |
| Bug Hunt Bench | Software › Software engineering | 25.7 | 31.1 | +5.4 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 37.6 | 46.3 | +4.3 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 60.8 | 57.5 | −3.3 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1629 | 1569 | −3.3 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 44.2 | 41.2 | −2.9 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 10.1 | 15.7 | +2.8 |
| Chartography (November 2026 rescoring) | Perception & media › Documents & charts | 84.4 | 87.2 | +2.8 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 85.6 | 88.1 | +2.5 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 90.4 | 92.5 | +2.1 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 22.0 | 24.0 | +2.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 76.7 | 78.4 | +1.8 |
| micro1 PharmaBench | Professional › General knowledge work | 9.7 | 11.4 | +1.7 |
| Epoch Capabilities Index (ECI) | Composite indexes | 165 | 167 | +1.7 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 63.0 | 61.3 | −1.7 |
| HiL-Bench (Human-in-Loop Benchmark) | Tool & computer use › General agents | 57.0 | 55.3 | −1.7 |
| OSWorld 2.1 Offline, 500-step Partial Score | Tool & computer use › Computer & browser use | 79.2 | 77.6 | −1.6 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1691 | 1703 | +1.5 |
| Legal Research Bench | Professional › Legal | 55.3 | 53.8 | −1.4 |
| OTIS Mock AIME 2024-2025 | Math | 98.9 | 100.0 | +1.1 |
| FrontierSWE V2 | Software › Software engineering | 52.0 | 53.1 | +1.1 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 97.5 | 98.5 | +1.0 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 88.4 | 87.8 | −0.6 |
| SWE-Marathon v1.1 | Software › Software engineering | 48.8 | 49.4 | +0.6 |
| FORTRESS Average Risk Score (ARS) | Safety & alignment | 16.8 | 17.5 | −0.6 |
| GBENCH One-shot Coding GScore | Games | 67.1 | 66.8 | −0.3 |
| BigFinanceBench private holdout (final-answer accuracy) | Professional › Finance, accounting & tax | 49.2 | 49.4 | +0.2 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.0 | 0.7 | −0.2 |
| LMArena Text Arena Elo | Language & writing › General assistance | 1490 | 1490 | 0.0 |
Data: Artificial Analysis, Epoch AI, Arena and 19 othersatlas.village.dev