Last week the Intelligence Index vs Cost Pareto frontier moved out substantially.

Artificial Analysis tracks large language models on a chart plotting a measured intelligence score against what a task costs to run, and periodically reports which models push that frontier outward, meaning they deliver more measured capability per dollar than any earlier model near their price point. In this post it names three models, Claude Fable 5.1, Muse Spark 1.3 and GPT-6 Astra, as each setting a new point on that frontier over the prior week.

The post's supporting chart was an animated GIF that could not be retrieved here, so the exact intelligence scores and prices behind the claim cannot be restated with confidence. What is established is only the outlet's summary claim that these three models each set a new efficient-intelligence point, not the size of the improvement.