Gemini 4 Neon vs GPT-6 Sol
Gemini 4 Neon leads by 5.4 capability points (≈95% interval +2.6 to +8.1): a clear lead over GPT-6 Sol · P(higher) ≈ 100%
Wins 41 of 46 shared benchmarks (4 lost, 1 tied)
Capability vs. relative cost
Gemini 4 NeonGPT-6 SolClosed weightsOpen weights
Benchmark wins41 – 4 · 46 shared, 1 tied
Comparing matching setups only (13 benchmarks excluded) · suites counted once
Area← Gemini 4 NeonGPT-6 Sol →
Data: Artificial Analysis, Vals AI, Epoch AI and 9 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Sol minus Gemini 4 Neon.
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 70.0 | 17.6 | −52.4 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 75.9 | 31.2 | −44.7 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 47.2 | 7.2 | −40.0 |
| EnterpriseBench: CoreCraft | Tool & computer use › Customer service & operations | 77.4 | 43.1 | −34.3 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 63.6 | 30.3 | −33.3 |
| Terminal-Bench Science 0.1 (Vals) | Science › Multidisciplinary | 62.9 | 30.0 | −32.9 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 50.5 | 19.5 | −31.0 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 89.2 | 58.3 | −30.8 |
| Mystery Game Puzzles | Games | 85.0 | 56.0 | −29.0 |
| Legal Research Bench | Professional › Legal | 56.3 | 28.8 | −27.4 |
| Chess Puzzles (task version 2.0) | Games | 83.0 | 56.0 | −27.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 79.8 | 53.0 | −26.7 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 71.3 | 45.2 | −26.1 |
| GDP.xlsx | Professional › General knowledge work | 43.1 | 17.1 | −26.0 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 18.4 | 6.5 | −23.3 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 56.3 | 33.2 | −23.1 |
| Harvey LAB (Mercor run, undisclosed ~213-task cohort) | Professional › Legal | 26.6 | 4.1 | −22.5 |
| SimpleBench | Knowledge & reasoning › General knowledge | 94.4 | 73.1 | −21.3 |
| APEX-Agents 1.1 | Professional › General knowledge work | 88.9 | 68.7 | −20.2 |
| Chartography | Perception & media › Documents & charts | 73.5 | 53.6 | −19.9 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1699 | 1359 | −18.7 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 79.3 | 61.7 | −17.6 |
| Vals Science Index | Science › Multidisciplinary | 62.6 | 45.0 | −17.6 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 44.5 | 9.8 | −17.4 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 61.4 | 44.2 | −17.2 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 54.8 | 21.4 | −16.7 |
| τ³-Banking (Mercor run, bm25_grep retrieval) | Tool & computer use › Customer service & operations | 59.5 | 43.0 | −16.5 |
| MysteryMechanism | Math | 45.0 | 30.2 | −14.9 |
| SimpleQA Verified (Epoch anti-abstention) | Knowledge & reasoning › General knowledge | 75.2 | 60.7 | −14.5 |
| Vending-Bench 2 | Business simulations | 16843 | 14428 | −14.3 |
| Terminal-Bench 4.0 | Software › Software engineering | 61.6 | 49.4 | −12.2 |
| APEX-Accounting 1.1 | Professional › Finance, accounting & tax | 70.8 | 59.7 | −11.1 |
| LMArena Text Arena Elo | Language & writing › General assistance | 1518 | 1458 | −10.6 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 44.4 | 35.6 | −8.8 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 33.4 | 25.2 | −8.2 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -6.0 | 10.0 | +8.0 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 19.2 | 3.9 | −7.7 |
| FrontierMath Tier 4 (v2) - Private | Math | 97.6 | 90.0 | −7.6 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 21.2 | 9.3 | −6.0 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 93.7 | 87.8 | −5.9 |
| BioMysteryBench (Vals Terminus 2) | Science › Life sciences | 80.0 | 74.8 | −5.2 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 78.0 | 80.9 | +2.9 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 31.7 | 33.9 | +2.2 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 91.2 | 89.8 | −1.4 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.0 | 1.2 | +0.1 |
| OTIS Mock AIME 2024-2025 | Math | 100.0 | 100.0 | 0.0 |
Data: Artificial Analysis, Vals AI, Epoch AI and 9 othersatlas.village.dev