A major upgrade focused on multi-step reasoning and 4x faster inference, now live in the API.
OpenAI released GPT-X today, positioning it as the new flagship for reasoning-heavy workloads. The model posts state-of-the-art results on GPQA, AIME, and SWE-bench Verified while running roughly 4x faster than o3 at half the price per output token. GPT-X is available in the API immediately and rolls out to ChatGPT Plus this week.