Fable-5.1-max takes fourth on eyebench-v3

Adonis Singh posted that Fable-5.1-max reached fourth place on eyebench-v3. He called the result the biggest jump he has seen from Anthropic on the benchmark. The post states the model trails only 5.6-terra-max by a narrow margin and 5.6-sol-max by a wider one. Singh, who tests large language models and contributes to MCBench, shared a screenshot of the leaderboard table ranking models by visual acuity. The tweet was retweeted by Teortaxes, a commentator focused on DeepSeek.