Gemini 3.8 Flash vs GPT-6.2 Sol
GPT-6.2 Sol leads by 6.3 capability points (≈95% interval +3.8 to +8.8): a clear lead over Gemini 3.8 Flash · P(higher) ≈ 100%
Wins 61 of 79 shared benchmarks (17 lost, 1 tied)
Capability vs. relative cost
Gemini 3.8 FlashGPT-6.2 SolClosed weightsOpen weights
Benchmark wins17 – 61 · 79 shared, 1 tied
Comparing matching setups only (10 benchmarks excluded) · suites counted once
Area← Gemini 3.8 FlashGPT-6.2 Sol →
Data: Artificial Analysis, Vals AI, Surge AI and 15 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Sol minus Gemini 3.8 Flash.
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 10.0 | 67.6 | +57.6 |
| AA-PersonalityBench Slop index | Language & writing › Style & personality | 71.4 | 15.4 | −56.6 |
| ARC-AGI-3 (Semi-Private) | Games | 35.0 | 90.1 | +55.1 |
| Terminal-Bench Science 0.1 (Vals) | Science › Multidisciplinary | 8.6 | 58.6 | +50.0 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 38.2 | 85.3 | +47.2 |
| AA-PersonalityBench Informativeness index | Language & writing › Style & personality | 81.0 | 37.7 | −43.7 |
| KradleBench | Games | 24.3 | 67.9 | +43.6 |
| Terminal-Bench 4.0 | Software › Software engineering | 19.7 | 60.6 | +40.9 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 19.7 | 58.6 | +38.9 |
| AA-3D v1.0 Elo | Software › Frontend & design | 974 | 1259 | +38.8 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 29.1 | 64.0 | +34.9 |
| DeepsecBench-AA (Artificial Analysis) | Software › Cybersecurity | 14.5 | 49.3 | +34.8 |
| Chartography | Perception & media › Documents & charts | 40.9 | 75.1 | +34.2 |
| AA-PersonalityBench Transparency index | Language & writing › Style & personality | 23.8 | 56.9 | +33.4 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 0.0 | 32.7 | +32.7 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 4.5 | 32.8 | +28.3 |
| AA-PersonalityBench Organization index | Language & writing › Style & personality | 73.3 | 45.7 | −27.9 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 30.8 | 57.3 | +26.4 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 6.0 | 32.0 | +26.0 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 10.6 | 36.1 | +25.5 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 47.0 | 72.4 | +25.4 |
| AA-PersonalityBench Expressiveness index | Language & writing › Style & personality | 79.1 | 55.3 | −24.0 |
| RiemannBench (November 2026 rescoring) | Math | 56.8 | 80.0 | +23.2 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 57.3 | 80.3 | +23.1 |
| EnterpriseBench: CoreCraft | Tool & computer use › Customer service & operations | 58.5 | 36.4 | −22.1 |
| Vals Science Index | Science › Multidisciplinary | 35.8 | 57.8 | +22.0 |
| VIALS-AA (pass^3) | Science › Life sciences | 41.0 | 62.7 | +21.7 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 15.6 | 37.2 | +21.7 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 6.0 | 27.6 | +21.6 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 18.7 | 40.2 | +21.5 |
| AA-PersonalityBench Imagination index | Language & writing › Style & personality | 41.0 | 60.1 | +19.3 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 1.5 | 11.3 | +19.1 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1204 | 1541 | +18.5 |
| Blueprint-Bench 2 | Perception & media › Spatial & visual reasoning | 0.4 | 0.6 | +18.4 |
| VIALS-AA (average of 3) | Science › Life sciences | 52.0 | 68.9 | +17.0 |
| BioMysteryBench (Vals Terminus 2) | Science › Life sciences | 62.2 | 77.8 | +15.6 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 41.6 | 56.8 | +15.2 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 2.0 | 17.2 | +15.2 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 65.1 | 77.7 | +12.6 |
| SimpleBench | Knowledge & reasoning › General knowledge | 82.4 | 69.9 | −12.5 |
| Artificial Analysis Cyber Index v1 | Software › Cybersecurity | 25.1 | 37.0 | +11.9 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 29.7 | 41.3 | +11.6 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 54.7 | 65.3 | +10.7 |
| VIALS-AA (pass@3) | Science › Life sciences | 64.0 | 74.5 | +10.6 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 60.9 | 71.3 | +10.4 |
| Vals Index v2.2 | Composite indexes | 56.0 | 66.1 | +10.1 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 4.0 | 23.3 | +9.6 |
| CWE-Bench-AA (Artificial Analysis) | Software › Cybersecurity | 52.5 | 61.7 | +9.2 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 78.7 | 87.6 | +9.0 |
| CyberGym-E2E-AA (Artificial Analysis) | Software › Cybersecurity | 8.4 | 0.0 | −8.4 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 23.7 | 40.4 | +8.3 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | -2.8 | 13.4 | +8.1 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 26.5 | 34.5 | +8.0 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 22.8 | 30.0 | +7.2 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 65.8 | 58.9 | −6.8 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 63.2 | 69.5 | +6.3 |
| AA-AnalystAgent (pass^5) | Professional › General knowledge work | 46.3 | 52.5 | +6.3 |
| AA-PersonalityBench Conviction index | Language & writing › Style & personality | 84.8 | 79.2 | −5.6 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 48.9 | 43.3 | −5.6 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 1.0 | 10.9 | +5.0 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 44.2 | 48.8 | +4.6 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 0.8 | 4.8 | +4.0 |
| GDP.xlsx | Professional › General knowledge work | 18.3 | 22.3 | +4.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 66.8 | 62.9 | −3.9 |
| Vals Finance Index | Professional › Finance, accounting & tax | 60.5 | 56.7 | −3.8 |
| Dual-Arm Sim (task completion) | Tool & computer use › Robotics | 0.0 | 3.8 | +3.8 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 89.2 | 85.8 | −3.3 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 5.5 | 2.3 | +3.2 |
| SWE Atlas v1.1 - Overall | Software › Software engineering | 57.3 | 60.5 | +3.1 |
| Vals Legal Index | Professional › Legal | 69.8 | 66.7 | −3.1 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 8.9 | 4.3 | −2.3 |
| Legal Research Bench | Professional › Legal | 38.9 | 37.0 | −1.9 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 72.9 | 74.3 | +1.4 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 98.5 | 97.5 | −1.0 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 83.2 | 84.2 | +1.0 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.0 | 1.2 | +0.6 |
| MysteryMechanism | Math | 36.5 | 36.9 | +0.5 |
| AA-PersonalityBench Warmth index | Language & writing › Style & personality | 60.0 | 60.1 | +0.1 |
| HLAB-Vals-1.1 | Professional › Legal | 12.1 | 12.1 | 0.0 |
Data: Artificial Analysis, Vals AI, Surge AI and 15 othersatlas.village.dev