Google DeepMind's model card lists scores for GPT-6.2 Sol across several benchmarks. Sol scored 75.1% on Chartography, 1.6 percentage points ahead of Gemini 4 Neon. On LVBench, Sol scored 87.6% and Neon scored 89.0%.
Tasks these results can inform: Use the public scores for GPT-6.2 Sol, Gemini 4 Neon, and Claude Opus 5.5 in the same model card to understand their relative results on the listed benchmarks.
Tasks these results should not be generalized to: Tasks not listed, real-world product experience, cross-benchmark rankings across different testing frameworks, or overall capability across all user scenarios.
Applicable model version: GPT-6.2 Sol as labeled in the table. The page provides no more specific model ID or revision number.
Test environment or client: The page does not specify the test environment or client for GPT-6.2 Sol.
Reasoning tier and parameters: Not specified on this page.
The model card says its evaluation covers agentic coding, knowledge work, computer use, multimodal understanding, science and math, and reasoning, and links to a separate methodology page. This page does not disclose the run parameters, sample counts, or complete harness for GPT-6.2 Sol's individual tests. The scores below are recorded only as shown alongside the other models in this model card. They are not an independent replication.
Chart understanding, Chartography: Sol scores 75.1%, above Neon's 73.5% and Opus 5.5's 66.3%.
LVBench: Sol scores 87.6%, below Neon's 89.0% and above Opus 5.5's 83.7%.
Long-context GraphWalks (256k to 1M, BFS F1): Sol scores 80.1%, below Neon's 95.2% and above Opus 5.5's 66.8%.
AutomationBench: Sol scores 41.3%, below Neon's 56.3% and slightly below Opus 5.5's 42.5%.
Agents' Last Exam v1.1: Sol scores 73.0%, below Neon's 75.8% and above Opus 5.5's 71.6%.
SWE-sweep and Mini-SWE use differently labeled harnesses. Sol scores 8.3% and 11.6%, respectively. These should not be combined into a single conclusion about coding ability.
The columns follow the model card's order for Gemini 4 Neon, Claude Opus 5.5, and GPT-6.2 Sol.
| Category | Benchmark (notes) | Neon | Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|---|
| Capability index | AA Index v5.0.1 | 61 | 59 | 57 |
| Capability index | Vals Index v2.2 | 70.9% | 70.1% | 66.1% |
| Agentic coding | Terminal-bench 4.1 (Cheating adjusted accuracy) | 65.2% | 65.8% | 61.5% |
| Agentic coding | FrontierCode 2.0 (Main score) | 46.2% | 47.3% | 43.8% |
| Agentic coding | SWE-sweep (Provider harness) | 39.0% | 10.1% | 8.3% |
| Agentic coding | Mini-SWE harness | 12.8% | 3.6% | 11.6% |
| Knowledge work | AutomationBench | 56.3% | 42.5% | 41.3% |
| Knowledge work | Agents' Last Exam v1.1 | 75.8% | 71.6% | 73.0% |
| Science and math | RiemannBench | 84.0% | 80.0% | 80.0% |
| Science and math | Terminal-Bench-Science 0.1 | 70.5% | 63.3% | 68.0% |
| Reasoning | HLE-Diamond | 71.3% | 67.9% | 64.0% |
| Multimodal understanding | MKA (Climb WSR) | 56.5% | 53.0% | 52.7% |
| Multimodal understanding | Chartography | 73.5% | 66.3% | 75.1% |
| Multimodal understanding | LVBench | 89.0% | 83.7% | 87.6% |
| Long context | GraphWalks (256k to 1M, BFS F1) | 95.2% | 66.8% | 80.1% |
The table presents scores across several distinct fields, and the results are mixed: Sol outperforms Opus 5.5 on some multimodal and knowledge-work items, but trails Neon on the capability indices, reasoning, and several coding benchmarks. Scores should be interpreted in light of each benchmark's definition, harness, and evaluation setup. The page does not provide enough configuration detail to reproduce Sol's results, nor does it give sample sizes or uncertainty intervals. These results therefore do not support a judgment of overall superiority or a claim of independent verification.
The original page links to the evaluation methodology for Gemini 4 Neon, but this collection covers only the model card. The model card itself does not list Sol's exact model identifier, reasoning tier, parameters, benchmark sample counts, or run artifacts. These scores cannot be reproduced from this page alone.
GPT-6.2 Sol