Gemini 3.8 Flash vs Gemini 4 Hydrogen
Gemini 4 Hydrogen leads by 5.0 capability points (≈95% interval +2.6 to +7.5): a clear lead over Gemini 3.8 Flash · P(higher) ≈ 100%
Wins 68 of 75 shared benchmarks (4 lost, 3 tied)
Capability vs. relative cost
Gemini 3.8 FlashGemini 4 HydrogenClosed weightsOpen weights
Benchmark wins4 – 68 · 75 shared, 3 tied
Comparing matching setups only (1 benchmark excluded) · suites counted once
Area← Gemini 3.8 FlashGemini 4 Hydrogen →
Data: Artificial Analysis, Vals AI, Surge AI and 20 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Hydrogen minus Gemini 3.8 Flash.
| FrontierMath Tier 4 (v2) - Private | Math | 22.0 | 87.8 | +65.9 |
| TxBench-Antibody-Development V1 (60 evaluations) | Science › Life sciences | 15.0 | 60.6 | +45.6 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 38.2 | 82.7 | +44.5 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 6.0 | 44.0 | +38.0 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 19.7 | 54.0 | +34.3 |
| Terminal-Bench 4.0 | Software › Software engineering | 19.7 | 54.0 | +34.3 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 4.5 | 38.5 | +34.0 |
| FrontierSWE V2 | Software › Software engineering | 19.6 | 53.1 | +33.4 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 10.0 | 41.4 | +31.4 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 29.1 | 58.8 | +29.7 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 6.0 | 33.6 | +27.6 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 31.7 | 57.5 | +25.8 |
| Chartography | Perception & media › Documents & charts | 40.9 | 66.6 | +25.7 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 47.0 | 70.8 | +23.8 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 1.5 | 13.5 | +23.5 |
| GDP.xlsx | Professional › General knowledge work | 18.3 | 40.9 | +22.6 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 2.0 | 24.0 | +22.0 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 41.6 | 63.4 | +21.8 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 29.7 | 51.1 | +21.5 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 10.6 | 31.7 | +21.1 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 57.3 | 77.0 | +19.7 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 68.4 | 88.1 | +19.6 |
| KradleBench | Games | 24.3 | 43.9 | +19.6 |
| TxBench Oligo Discovery V1 (113 evaluations) | Science › Life sciences | 35.1 | 54.3 | +19.2 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 26.5 | 44.6 | +18.1 |
| Mystery Game Puzzles | Games | 47.0 | 65.0 | +18.0 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 18.7 | 36.3 | +17.6 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1204 | 1497 | +16.0 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1585 | 1703 | +15.7 |
| Legal Research Bench | Professional › Legal | 38.9 | 53.8 | +14.9 |
| HiL-Bench (Human-in-Loop Benchmark) | Tool & computer use › General agents | 41.5 | 55.3 | +13.9 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 60.9 | 74.4 | +13.5 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 65.1 | 78.0 | +12.9 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 4.0 | 29.3 | +12.6 |
| Vals Index v2.2 | Composite indexes | 56.0 | 68.5 | +12.5 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 66.8 | 78.4 | +11.7 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 15.6 | 27.2 | +11.7 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 23.7 | 46.3 | +11.3 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 63.2 | 73.7 | +10.5 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 30.8 | 41.2 | +10.4 |
| GBENCH One-shot Coding GScore | Games | 56.6 | 66.8 | +10.2 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | -2.8 | 17.4 | +10.1 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 44.2 | 54.2 | +10.0 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 22.8 | 32.8 | +10.0 |
| Bug Hunt Bench | Software › Software engineering | 17.1 | 27.0 | +9.8 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 0.0 | 9.7 | +9.7 |
| SimpleBench | Knowledge & reasoning › General knowledge | 82.4 | 91.7 | +9.3 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 78.7 | 87.8 | +9.1 |
| TxBench Antibody Discovery V1 (100 evaluations) | Science › Life sciences | 49.0 | 57.7 | +8.7 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 1.0 | 15.7 | +7.4 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 8.9 | -4.5 | −6.7 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 54.7 | 61.3 | +6.7 |
| Epoch Capabilities Index (ECI) | Composite indexes | 159 | 167 | +6.4 |
| micro1 PharmaBench | Professional › General knowledge work | 5.6 | 11.4 | +5.8 |
| Vals Finance Index | Professional › Finance, accounting & tax | 60.5 | 66.1 | +5.7 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 5.5 | 0.0 | +5.5 |
| FORTRESS Average Risk Score (ARS) | Safety & alignment | 22.7 | 17.5 | +5.3 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 48.9 | 53.1 | +4.2 |
| Vals Legal Index | Professional › Legal | 69.8 | 73.6 | +3.8 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 89.2 | 92.5 | +3.3 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 72.9 | 75.8 | +2.9 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 0.8 | 3.6 | +2.8 |
| Professional Reasoning Benchmark - Finance | Professional › Finance, accounting & tax | 49.5 | 52.3 | +2.8 |
| BuseyBench SVG | Perception & media › Generation | 6.6 | 6.9 | +2.3 |
| Chess Puzzles (task version 2.0) | Games | 58.0 | 60.0 | +2.0 |
| LMArena Text Arena Elo | Language & writing › General assistance | 1497 | 1490 | −1.2 |
| SimpleQA Verified (Epoch anti-abstention) | Knowledge & reasoning › General knowledge | 69.7 | 70.9 | +1.2 |
| OTIS Mock AIME 2024-2025 | Math | 98.9 | 100.0 | +1.1 |
| HLAB-Vals-1.1 | Professional › Legal | 12.1 | 11.3 | −0.8 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 65.8 | 65.1 | −0.6 |
| GradeBench | Professional › Education | 51.5 | 51.9 | +0.4 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.0 | 0.7 | +0.3 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 98.5 | 98.5 | 0.0 |
| Dual-Arm Sim (task completion) | Tool & computer use › Robotics | 0.0 | 0.0 | 0.0 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 83.2 | 83.2 | 0.0 |
Data: Artificial Analysis, Vals AI, Surge AI and 20 othersatlas.village.dev