Gemini 4 Carbon vs GPT-6.2 Sol
Gemini 4 Carbon leads by 0.8 capability points (≈95% interval −2.0 to +3.5): statistically tied · P(higher) ≈ 71%
Wins 52 of 70 shared benchmarks (16 lost, 2 tied)
Capability vs. relative cost
Gemini 4 CarbonGPT-6.2 SolClosed weightsOpen weights
Benchmark wins52 – 16 · 70 shared, 2 tied
Comparing matching setups only (38 benchmarks excluded) · suites counted once
Area← Gemini 4 CarbonGPT-6.2 Sol →
Data: Artificial Analysis, Surge AI, Vals 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 4 Carbon.
| KradleBench | Games | 22.0 | 67.9 | +46.0 |
| EnterpriseBench: CoreCraft | Tool & computer use › Customer service & operations | 80.5 | 36.4 | −44.1 |
| SimpleBench | Knowledge & reasoning › General knowledge | 92.6 | 69.9 | −22.7 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 13.8 | 32.7 | +18.9 |
| Legal Research Bench | Professional › Legal | 55.8 | 37.0 | −18.8 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 35.2 | 17.2 | −18.0 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 19.4 | 37.2 | +17.8 |
| GDP.xlsx | Professional › General knowledge work | 38.9 | 22.3 | −16.6 |
| GraphWalks BFS: 256K to 1M Aggregate | Knowledge & reasoning › Reading & long context | 96.2 | 80.1 | −16.1 |
| HANDBOOK.md Agents v1.1 | Tool & computer use › Customer service & operations | 49.8 | 34.5 | −15.3 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 71.1 | 56.8 | −14.3 |
| MultiChallenge | Language & writing › Conversation | 85.9 | 72.0 | −13.9 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 76.4 | 62.9 | −13.5 |
| DeepsecBench — Revalidated V1 (233 issues) | Software › Cybersecurity | 35.2 | 47.3 | +12.2 |
| AutomationBench 1.0.6 (Zapier) | Tool & computer use › Customer service & operations | 53.0 | 41.3 | −11.7 |
| SWE Atlas v1.1 - Overall | Software › Software engineering | 71.5 | 60.5 | −11.0 |
| Vals Finance Index | Professional › Finance, accounting & tax | 67.0 | 56.7 | −10.4 |
| Professional Reasoning Benchmark - Legal | Professional › Legal | 53.4 | 43.3 | −10.0 |
| Dual-Arm Sim (task completion) | Tool & computer use › Robotics | 13.8 | 3.8 | −10.0 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 16.1 | 11.3 | −9.3 |
| DrugDiscoveryBench | Science › Life sciences | 71.5 | 62.2 | −9.3 |
| Vals Legal Index | Professional › Legal | 75.9 | 66.7 | −9.3 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 23.8 | 32.8 | +9.0 |
| ARC-AGI-2 | Knowledge & reasoning › Abstract reasoning | 94.2 | 85.8 | −8.3 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 67.2 | 58.9 | −8.2 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 39.4 | 23.3 | −8.1 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 56.4 | 48.8 | −7.6 |
| Artificial Analysis Legal Index | Professional › Legal | 62.0 | 55.0 | −6.9 |
| ARC-AGI-3 (Semi-Private) | Games | 96.9 | 90.1 | −6.8 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 53.5 | 40.4 | −6.6 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 77.8 | 71.3 | −6.5 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 30.6 | 36.1 | +5.5 |
| HLAB-Vals-1.1 | Professional › Legal | 17.5 | 12.1 | −5.4 |
| Artificial Analysis Finance & Accounting Index | Professional › Finance, accounting & tax | 62.6 | 57.2 | −5.4 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 80.1 | 85.3 | +5.3 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 69.1 | 64.0 | −5.1 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -5.7 | 4.3 | +5.0 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 60.6 | 56.1 | −4.5 |
| Agents' Last Exam v1.1 Score | Professional › General knowledge work | 77.4 | 73.0 | −4.4 |
| Artificial Analysis Engineering Index | Science › Engineering | 65.3 | 61.2 | −4.1 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 82.3 | 78.3 | −4.0 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1611 | 1541 | −3.8 |
| VIALS-AA (pass@3) | Science › Life sciences | 70.8 | 74.5 | +3.7 |
| Artificial Analysis Healthcare & Medical Index | Health & medicine | 56.8 | 53.3 | −3.5 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 17.2 | 10.9 | −3.1 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 83.3 | 80.3 | −3.0 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 60.0 | 57.3 | −2.8 |
| Chartography | Perception & media › Documents & charts | 72.5 | 75.1 | +2.6 |
| Terminal-Bench 4.0 | Software › Software engineering | 63.1 | 60.6 | −2.5 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 25.2 | 27.6 | +2.4 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 17.9 | 13.4 | −2.3 |
| Artificial Analysis Economics Index | Composite indexes | 59.6 | 57.4 | −2.3 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 76.5 | 74.3 | −2.2 |
| SWE-sweep | Software › Software engineering | 9.6 | 11.6 | +2.0 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 39.5 | 41.4 | +1.9 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 62.4 | 60.5 | −1.9 |
| GDPval v2 (OpenAI internal) | Professional › General knowledge work | 48.8 | 46.9 | −1.9 |
| VIALS-AA (average of 3) | Science › Life sciences | 67.3 | 68.9 | +1.7 |
| Blueprint-Bench 2 | Perception & media › Spatial & visual reasoning | 0.6 | 0.6 | −1.6 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 74.0 | 72.4 | −1.6 |
| Artificial Analysis Strategy & Ops Index | Professional › General knowledge work | 60.9 | 59.4 | −1.5 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 89.1 | 87.6 | −1.4 |
| ComplexConstraints (November 2026 rescoring) | Language & writing › Instruction following | 70.9 | 69.5 | −1.4 |
| VIALS-AA (pass^3) | Science › Life sciences | 64.0 | 62.7 | −1.2 |
| ARC-AGI-1 Semi-Private | Knowledge & reasoning › Abstract reasoning | 98.5 | 97.5 | −1.0 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 66.0 | 65.3 | −0.7 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 29.6 | 30.0 | +0.4 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.8 | 1.2 | +0.2 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 4.8 | 4.8 | 0.0 |
| RiemannBench (November 2026 rescoring) | Math | 80.0 | 80.0 | 0.0 |
Data: Artificial Analysis, Surge AI, Vals AI and 15 othersatlas.village.dev