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 quietly updated its evaluation metrics for the GPT-6 Astra model, making changes that appear to favor Astra. These revisions to several evaluation benchmarks have occurred since the initial blog post announcement on September 3. The company continues to revise other metrics even after the model's launch, as reported by Emily Forlini for Fortune.