Outside testers scored OpenAI's new Astra below Anthropic's Fable 5.1.
Artificial Analysis, an independent benchmarking firm, published its Astra results on 4 September. Astra scored 61 on their intelligence index, five points under Claude Fable 5.1 and level with OpenAI's own Sol. On coding agents it reached 67 against Fable's 70, while costing 2.5 times its predecessor.
Why it mattersLaunch-day claims come from the lab, and the ranking that matters comes days later from outsiders. If you pay per token, Astra is the expensive option for coding right now.
Read in the app — free, in 9 languages
Related stories
A Chinese lab built Saudi Arabia's flagship Arabic AI model.2026-09-04CrowdStrike built an attack AI that hunts its own defence AI.
2026-09-04Attackers are stealing OpenAI keys from a popular AI building tool.
2026-09-04A startup's AI found six curl flaws after OpenAI's found none.
2026-09-04Google's cheapest new model doubles in price on 1 January.
2026-09-04