OpenAI quietly updates its evaluation metrics for GPT-6 Astra, making changes that appear to favor Astra and continuing to revise other metrics after launch (Emily Forlini/Fortune)

OpenAI has updated its evaluation metrics for the GPT-6 Astra model, making changes that appear to favor Astra. According to Fortune, several evaluation benchmarks for GPT-6 Astra have been changed since the model's launch was announced on September 3.

The GPT-6 Astra model has shown significant improvement across benchmark testing and includes new business features. However, concerns have been raised about its new 'recurrent depth' reasoning capabilities, which allow the model to consider a problem multiple times before taking an action. This is different from the standard chain-of-thought reasoning used in previous models.

OpenAI chief scientist Jakub Pachocki stated that the company will not accept degradation in its ability to monitor model alignment beyond a certain level and will withhold scaling until it can regain enough confidence. TechRadar reports that numerous experts have expressed concerns about the model's new reasoning architecture.

Brief written by urgent.news from Techmeme, TechRadar — 2 reports on this story. Machine-written — may contain errors; check the original before relying on it.