Part 3/8:
This enhanced model offers a 34% reduction in major errors on challenging tasks, making it a much more reliable solution for users. Notably, OpenAI's reasoning researcher Nom Brown showcased the AI's capabilities by generating a three-paragraph essay about strawberries without including the letter "e." While performance across the board has notably improved, there are reported reductions in some advanced benchmarks, leading to questions regarding the effectiveness of earlier builds in testing.
Beginning with subscribers in the Plus and Team tiers, the updated model will roll out to Enterprise and Education users next week, ensuring a broad reach across various applications.