Gemini 4 Hydrogen vs Muse Ember
TryTop 2 now
Gemini 4 Hydrogen leads by 3.5 capability points (≈95% interval +1.0 to +6.1): a clear lead over Muse Ember · P(higher) ≈ 100%
Wins 45 of 53 shared benchmarks (8 lost)
Capability vs. relative cost
Gemini 4 HydrogenMuse EmberClosed weightsOpen weights
Benchmark wins45 – 8 · 53 shared
Comparing matching setups only (1 benchmark excluded) · suites counted once
Area← Gemini 4 HydrogenMuse Ember →
Data: Artificial Analysis, Vals AI, Arena and 7 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Ember minus Gemini 4 Hydrogen.
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 44.0 | 2.0 | −42.0 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 82.7 | 47.1 | −35.6 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 58.8 | 25.1 | −33.7 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 49.5 | 17.6 | −31.9 |
| GDP.xlsx | Professional › General knowledge work | 40.9 | 15.4 | −25.5 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 13.5 | 1.0 | −24.5 |
| Ramp SWE-Bench v2 | Software › Software engineering | 71.6 | 51.8 | −19.8 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 46.3 | 7.7 | −19.3 |
| SimpleBench | Knowledge & reasoning › General knowledge | 91.7 | 72.7 | −19.0 |
| HiL-Bench (Human-in-Loop Benchmark) | Tool & computer use › General agents | 55.3 | 74.0 | +18.7 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 80.6 | 62.0 | −18.6 |
| Artificial Analysis Engineering Index | Science › Engineering | 60.1 | 41.8 | −18.3 |
| Artificial Analysis Legal Index | Professional › Legal | 56.7 | 38.5 | −18.3 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 53.5 | 35.4 | −18.2 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 39.0 | 22.0 | −17.0 |
| DrugDiscoveryBench | Science › Life sciences | 69.1 | 52.4 | −16.7 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 29.3 | -2.7 | −16.0 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 73.7 | 58.8 | −14.9 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 44.6 | 29.8 | −14.8 |
| Terminal-Bench 4.0 | Software › Software engineering | 54.0 | 39.4 | −14.6 |
| BuseyBench SVG | Perception & media › Generation | 6.9 | 5.4 | −14.6 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 38.5 | 24.2 | −14.3 |
| Artificial Analysis Economics Index | Composite indexes | 55.0 | 41.0 | −14.1 |
| HLAB-Vals-1.1 | Professional › Legal | 11.3 | 25.0 | +13.8 |
| Artificial Analysis Finance & Accounting Index | Professional › Finance, accounting & tax | 58.1 | 44.6 | −13.5 |
| Artificial Analysis Coding Agent Index v2.1 | Software › Software engineering | 52.3 | 39.7 | −12.6 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1703 | 1608 | −12.6 |
| Artificial Analysis Healthcare & Medical Index | Health & medicine | 53.3 | 41.3 | −12.1 |
| GBENCH One-shot Coding GScore | Games | 66.8 | 56.4 | −10.4 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 32.8 | 22.8 | −10.0 |
| Artificial Analysis Strategy & Ops Index | Professional › General knowledge work | 57.7 | 48.7 | −9.1 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 27.2 | 18.3 | −8.9 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 15.7 | -2.0 | −8.9 |
| Vals Index v2.2 | Composite indexes | 68.5 | 60.5 | −8.0 |
| Chartography (November 2026 rescoring) | Perception & media › Documents & charts | 87.2 | 79.3 | −7.9 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 78.4 | 71.4 | −7.1 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 24.0 | 17.2 | −6.8 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 74.4 | 68.1 | −6.4 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 17.4 | 4.8 | −6.3 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -4.5 | 6.6 | +5.6 |
| FORTRESS Average Risk Score (ARS) | Safety & alignment | 17.5 | 11.9 | +5.5 |
| Vals Finance Index | Professional › Finance, accounting & tax | 66.1 | 60.9 | −5.2 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 41.2 | 36.0 | −5.2 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 54.2 | 49.2 | −5.0 |
| Legal Research Bench | Professional › Legal | 53.8 | 49.0 | −4.8 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 87.8 | 83.7 | −4.1 |
| Professional Reasoning Benchmark - Finance | Professional › Finance, accounting & tax | 52.3 | 56.0 | +3.7 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 75.8 | 72.2 | −3.6 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 53.1 | 55.6 | +2.5 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1569 | 1526 | −2.3 |
| Vals Legal Index | Professional › Legal | 73.6 | 75.7 | +2.1 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 65.1 | 63.8 | −1.3 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.7 | 0.9 | +0.1 |
Data: Artificial Analysis, Vals AI, Arena and 7 othersatlas.village.dev