Artificial Analysis’s snapshot shows an Intelligence Index score of 41–57 for Sol, from Low to Max. Higher levels take longer per index task and have slightly higher output speeds, while the Max overview reports a per-task cost of $1.01.
Tasks this is suitable for assessing: Comparing the five levels tested on this page by their Artificial Analysis Intelligence Index scores, task times, output speeds, and task costs.
Tasks this should not be generalized to: Tasks not included in this index, other testing platforms, or latency and actual bills in production environments.
Applicable model version: GPT-6.2 Sol. Levels are Low, Medium, High, Xhigh, and Max.
Test environment or client: Artificial Analysis independent evaluation. The page notes that API Provider Benchmarks data is not public.
Reasoning levels and parameters: The five levels are listed above. The model’s default effort is medium. The page does not provide other reasoning parameters.
Intelligence Index v5.0.1 aggregates nine evaluations: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1. Higher index scores are better.
The page defines the time per index task as the number of output tokens divided by end-to-end output speed. Output speed is calculated using the total request time, including the wait for the first token. Time to First Answer Token separately measures the seconds before the first answer token and includes the thinking time of reasoning models. Task cost is the weighted average cost of completing the index tasks.
| Reasoning level | Index score | Time per index task | Output speed |
|---|---|---|---|
| Low | 41 | 1.5 minutes | 65 tok/s |
| Medium | 53 | 3.3 minutes | 69 tok/s |
| High | 55 | 4.7 minutes | 72 tok/s |
| Xhigh | 56 | 5.4 minutes | 75 tok/s |
| Max | 57 | 8.7 minutes | 77 tok/s |
In this snapshot, Low and Max differ by 16 points. Max takes about 5.8 times as long per task as Low. The overview card also reports a time to first answer token of 76.14 seconds for Max, an input price of $2 per million tokens, an output price of $10, a 95% cache discount, and a cost of $1.01 per index task.
The Max overview on the page gives an Intelligence Index score of 57 (ranked 13/96), 8.7 minutes per index task, an output speed of 77 tok/s, and 76.14 seconds to the first answer token.
The comparison table shows a 1M context window for all five levels. The Price column shows $1.47 for each. This is a different metric from the $1.01 per-task cost in the overview, and the page excerpt does not explain the pricing basis for $1.47.
The page summary says Max generated 60.9M output tokens in this index. The index task cost chart is a comparison view of 28/763 models.
The page lists “259 models in this class” but does not explain how the class is defined. The index leaderboard view separately shows 28/763 models. These should not be treated as the same comparison sample.
These data support comparing the five levels under this index and the page’s timing method: scores rise with the level, but time per task increases. Output speed changes by less. They do not support inferring a fixed latency for general requests or treating the per-task cost as a quote for an arbitrary real-world request. The page is a dynamic leaderboard, and this record reflects only the snapshot observed on 2026-11-19. The page does not show complete raw responses for each level or retest variance.
Visit the original model page, select each reasoning level for GPT-6.2 Sol, and compare the Intelligence, Time per Task, Output Speed, and Cost metrics. The page says API Provider Benchmarks data is not public, so the measured results on the provider API side cannot be reproduced from this page alone.
GPT-6.2 Sol