OpenAI has replaced the default ChatGPT model with GPT-5.5 Instant, a lighter variant of the flagship GPT-5.5 model launched in April. The new default emphasizes factual accuracy, shorter responses, and closer instruction-following for everyday tasks like writing and analysis. Benchmark improvements include a jump from 75.0% to 81.6% on the CharXiv scientific chart reasoning benchmark and from 69.2% to 76.0% on MMMU-Pro. OpenAI is also introducing 'memory sources' across all ChatGPT models, letting users see which personal data inputs informed a response and remove them. GPT-5.5 Instant competes in the same fast, low-cost model tier as Google's Gemini Flash and Anthropic's Claude Haiku.
Sort: