benchmark-discrepancy

Tag

Cards List
#benchmark-discrepancy

OpenAI's AGI number came from a harness, not the model (6 minute read)

TLDR AI · 6d ago Cached

OpenAI's reported 99.9% AGI benchmark score for GPT-6 Astra was achieved using a specific harness (Provider Adapter), while the standard harness yields a much lower 62.7% score, highlighting significant issues in AI model evaluation and benchmarking transparency.

0 favorites 0 likes
← Back to home

Submit Feedback