What the snapshot shows
Anthropic's frontier model, Claude Fable 5, records a 1507.5 LMArena rating. Meta follows at 1491.6.
The published 95% confidence interval for Anthropic is 1500.0 to 1515.0. The interval for Meta is 1482.5 to 1500.7. Those intervals overlap, so the point-estimate lead should not be read as a settled margin.
How to read the result
This is one source view of human preference within a specific LMArena subset. It is not a Model Gauntlet consensus score and it does not establish the best model for every task.
The current snapshot remains separate from the selected historical archive. Model Gauntlet does not calculate movement between those incompatible measurement windows.