Claude Fable 5.1 vs Gemini 4 Hydrogen
Gemini 4 Hydrogen leads by 0.1 capability points (≈95% interval −2.4 to +2.7): statistically tied · P(higher) ≈ 54%
Wins 31 of 58 shared benchmarks (24 lost, 3 tied)
Capability vs. relative cost
Claude Fable 5.1Gemini 4 HydrogenClosed weightsOpen weights
Benchmark wins24 – 31 · 58 shared, 3 tied
Comparing matching setups only (5 benchmarks excluded) · suites counted once
Area← Claude Fable 5.1Gemini 4 Hydrogen →
Data: Surge AI, Epoch AI, Arena and 21 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Hydrogen minus Claude Fable 5.1.
| TxBench-Antibody-Development V1 (60 evaluations) | Science › Life sciences | 27.8 | 60.6 | +32.8 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 22.4 | 51.1 | +28.8 |
| Chartography | Perception & media › Documents & charts | 46.2 | 66.6 | +20.4 |
| GDP.xlsx | Professional › General knowledge work | 23.1 | 40.9 | +17.8 |
| EBR-bench v4 (card ban, single agent) | Games | 57.1 | 70.5 | +13.3 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 60.4 | 73.7 | +13.3 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 22.5 | 9.7 | −12.8 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 70.0 | 57.5 | −12.5 |
| Chess Puzzles (task version 2.0) | Games | 48.0 | 60.0 | +12.0 |
| Bug Hunt Bench | Software › Software engineering | 41.0 | 31.1 | −9.8 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 18.3 | 27.2 | +8.9 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 31.2 | 24.0 | −7.2 |
| Mystery Game Puzzles | Games | 58.0 | 65.0 | +7.0 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 63.8 | 70.8 | +7.0 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 84.0 | 77.0 | −7.0 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 9.3 | -4.5 | −6.9 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 27.2 | 33.6 | +6.4 |
| HiL-Bench (Human-in-Loop Benchmark) | Tool & computer use › General agents | 61.5 | 55.3 | −6.2 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 9.6 | 3.6 | −6.0 |
| SpireClimb | Games | 2.0 | 7.7 | +5.7 |
| FrontierCode 2.0 - Pass Rate | Software › Software engineering | 37.6 | 43.2 | +5.6 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1741 | 1703 | −5.1 |
| SimpleBench | Knowledge & reasoning › General knowledge | 86.6 | 91.7 | +5.1 |
| FrontierCode 2.0 - Score | Software › Software engineering | 34.0 | 38.9 | +4.8 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 49.4 | 54.2 | +4.8 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1577 | 1497 | −4.4 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 76.6 | 80.6 | +4.0 |
| SWE-Marathon v1.1 | Software › Software engineering | 45.6 | 49.4 | +3.8 |
| FORTRESS Average Risk Score (ARS) | Safety & alignment | 13.7 | 17.5 | −3.7 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 90.0 | 92.5 | +2.5 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 47.1 | 44.6 | −2.5 |
| CursorBench 4.0 | Software › Software engineering | 51.8 | 54.3 | +2.5 |
| LMArena Text Arena Elo | Language & writing › General assistance | 1502 | 1490 | −2.2 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 90.2 | 88.1 | −2.1 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 36.5 | 38.5 | +2.0 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 42.0 | 44.0 | +2.0 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 13.5 | 17.4 | +2.0 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 85.1 | 83.2 | −1.9 |
| BuseyBench SVG | Perception & media › Generation | 7.0 | 6.9 | −1.7 |
| Professional Reasoning Benchmark - Finance | Professional › Finance, accounting & tax | 50.8 | 52.3 | +1.5 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 51.6 | 53.1 | +1.5 |
| micro1 PharmaBench | Professional › General knowledge work | 12.8 | 11.4 | −1.4 |
| GradeBench | Professional › Education | 53.1 | 51.9 | −1.2 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 97.5 | 98.5 | +1.0 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 1.0 | 0.0 | +1.0 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 76.7 | 75.8 | −0.9 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 14.2 | 15.7 | +0.8 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 77.9 | 78.4 | +0.6 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 28.1 | 29.3 | +0.6 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 32.2 | 31.7 | −0.5 |
| GBENCH One-shot Coding GScore | Games | 67.3 | 66.8 | −0.5 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 13.7 | 13.5 | −0.3 |
| Epoch Capabilities Index (ECI) | Composite indexes | 167 | 167 | −0.2 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.0 | 0.7 | −0.1 |
| SimpleQA Verified (Epoch anti-abstention) | Knowledge & reasoning › General knowledge | 70.8 | 70.9 | +0.1 |
| Dual-Arm Sim (task completion) | Tool & computer use › Robotics | 0.0 | 0.0 | 0.0 |
| FrontierMath Tier 4 (v2) - Private | Math | 87.8 | 87.8 | 0.0 |
| OTIS Mock AIME 2024-2025 | Math | 100.0 | 100.0 | 0.0 |
Data: Surge AI, Epoch AI, Arena and 21 othersatlas.village.dev