Frontier benchmarks move three times in seven days as Astra and Fable 5.1 tie for first
The author notes that Fable 5.1 and GPT-6 Astra launched in quick succession and the Intelligence Index shipped two new versions, leaving leaderboard-based model picks with data that lasts only days.
Frontier Benchmarks Moved Three Times in Seven Days. The Rankings Flipped With Them. Fable 5.1 launched. GPT-6 Astra launched. Artificial Analysis re-shipped the Intelligence Index to v4.2, then v4.3. By Friday, Fable 5.1 and Astra were tied for first. If your team is picking models from a leaderboard, you are picking on data with a shelf life of days. The only routing intelligence that survives the churn is the one your workspace measures on your own work. https://tilted.xyz/blog/the-benchmarks-just-moved-three-times-in-seven-days-your-team-cannot-pick-models-from-a-leaderboard-anymore