GPT-6.1 Terra vs GPT-6.2 Sol
GPT-6.2 Sol leads by 1.7 capability points (≈95% interval −0.9 to +4.3): statistically tied · P(higher) ≈ 90%
Wins 54 of 63 shared benchmarks (8 lost, 1 tied)
Capability vs. relative cost
GPT-6.1 TerraGPT-6.2 SolClosed weightsOpen weights
Benchmark wins8 – 54 · 63 shared, 1 tied
Comparing matching setups only (2 benchmarks excluded) · suites counted once
Area← GPT-6.1 TerraGPT-6.2 Sol →
Data: Artificial Analysis, Vals AI, Arena 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 GPT-6.1 Terra.
| KernelGen1P - OpenAI release percentage | Software › Data & ML | 45.3 | 74.7 | +53.9 |
| FrontierCUA v1 pass@1 | Tool & computer use › Computer & browser use | 7.5 | 25.0 | +17.5 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 47.3 | 64.0 | +16.7 |
| PostTrainBench Lite - release-reported performance | Software › Data & ML | 60.5 | 44.9 | −15.6 |
| Blueprint-Bench 2 | Perception & media › Spatial & visual reasoning | 0.4 | 0.6 | +14.9 |
| Bug Hunt Bench | Software › Software engineering | 38.7 | 52.1 | +13.3 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 72.2 | 85.3 | +13.2 |
| Chartography | Perception & media › Documents & charts | 63.2 | 75.1 | +11.9 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1400 | 1612 | +11.6 |
| Terminal-Bench 4.0 | Software › Software engineering | 49.5 | 60.6 | +11.1 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 60.2 | 71.3 | +11.1 |
| Legal Research Bench | Professional › Legal | 26.0 | 37.0 | +11.1 |
| RWE-bench | Science › Engineering | 19.0 | 30.0 | +11.0 |
| HLAB-Vals-1.1 | Professional › Legal | 1.7 | 12.1 | +10.4 |
| Terminal-Bench-Science 0.1 | Science › Multidisciplinary | 57.6 | 67.6 | +10.0 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1682 | 1757 | +9.9 |
| TxBench Preclinical Pharmacology V1 Clean Reruns | Science › Life sciences | 55.7 | 65.3 | +9.7 |
| Artificial Analysis Strategy & Ops Index | Professional › General knowledge work | 50.0 | 59.4 | +9.4 |
| Artificial Analysis Economics Index | Composite indexes | 48.5 | 57.4 | +8.9 |
| Artificial Analysis Legal Index | Professional › Legal | 46.4 | 55.0 | +8.6 |
| Artificial Analysis Healthcare & Medical Index | Health & medicine | 44.7 | 53.3 | +8.6 |
| Artificial Analysis Finance & Accounting Index | Professional › Finance, accounting & tax | 49.0 | 57.2 | +8.2 |
| TaxCalcBench TY25 | Professional › Finance, accounting & tax | 24.0 | 32.0 | +8.0 |
| Vals Legal Index | Professional › Legal | 58.8 | 66.7 | +7.9 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 55.0 | 62.9 | +7.9 |
| Medical Long Context Reasoning (MLCR-AA, October 2026 methodology) | Health & medicine | 30.6 | 37.2 | +6.7 |
| Artificial Analysis Engineering Index | Science › Engineering | 54.7 | 61.2 | +6.4 |
| CWE-bench v2 Pass@1 | Software › Cybersecurity | 60.0 | 66.0 | +6.0 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 78.2 | 84.2 | +6.0 |
| Artificial Analysis Intelligence Index v5.0.1 | Composite indexes | 51.3 | 57.3 | +5.9 |
| Vals Index v2.2 | Composite indexes | 60.5 | 66.1 | +5.5 |
| AA-Omniscience v1.1 Index | Knowledge & reasoning › General knowledge | 29.6 | 40.4 | +5.4 |
| Gray Swan IPI Attack Success (k=15) | Safety & alignment | 7.3 | 2.3 | +5.0 |
| MentalHealthBench | Health & medicine | 55.5 | 50.9 | −4.6 |
| Terminal-Bench 4.0 (Artificial Analysis) | Software › Software engineering | 55.6 | 60.1 | +4.5 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | 12.7 | 4.3 | −4.2 |
| DrugDiscoveryBench | Science › Life sciences | 58.1 | 62.2 | +4.1 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 84.0 | 87.6 | +3.6 |
| Vals Finance Index | Professional › Finance, accounting & tax | 53.3 | 56.7 | +3.3 |
| VIALS-AA (pass@3) | Science › Life sciences | 71.4 | 74.5 | +3.1 |
| Ramp SWE-Bench v2 | Software › Software engineering | 70.4 | 73.4 | +3.0 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 56.0 | 58.9 | +3.0 |
| ExploitBench - release-reported exploit progress | Software › Cybersecurity | 97.1 | 100.0 | +2.9 |
| VIALS-AA (average of 3) | Science › Life sciences | 66.0 | 68.9 | +2.9 |
| CritPt v2 (agentic, avg@5) | Science › Physical sciences | 81.4 | 78.6 | −2.9 |
| SWE Atlas v1.1 - Overall | Software › Software engineering | 57.7 | 60.5 | +2.8 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 17.9 | 23.3 | +2.7 |
| VIALS-AA (pass^3) | Science › Life sciences | 60.2 | 62.7 | +2.5 |
| GPQA Diamond | Science › Multidisciplinary | 93.6 | 96.0 | +2.4 |
| SimpleBench | Knowledge & reasoning › General knowledge | 72.2 | 69.9 | −2.3 |
| MultiChallenge | Language & writing › Conversation | 74.2 | 72.0 | −2.2 |
| FrontierCode 2.0 (Artificial Analysis) | Software › Software engineering | 42.3 | 44.4 | +2.1 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 78.6 | 80.3 | +1.7 |
| OpenAI research debugging - internal release score | Software › Data & ML | 76.4 | 74.7 | −1.7 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 10.5 | 11.3 | +1.6 |
| GDP.pdf (Artificial Analysis, All-pass) | Professional › General knowledge work | 28.6 | 30.0 | +1.4 |
| HealthBench Professional | Health & medicine | 64.2 | 65.5 | +1.3 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 8.5 | 10.9 | +1.2 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 73.3 | 74.3 | +1.0 |
| Agents' Last Exam v1.1 (ALE v1.1) | Professional › General knowledge work | 41.0 | 41.8 | +0.8 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 13.2 | 13.4 | +0.1 |
| Artificial Analysis Coding Agent Index v2.1 | Software › Software engineering | 48.6 | 48.6 | 0.0 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 1.2 | 1.2 | 0.0 |
Data: Artificial Analysis, Vals AI, Arena and 15 othersatlas.village.dev