OpenAI has released GPT-5.5 Instant as the default model for ChatGPT, aiming to significantly reduce factual errors and hallucinations. Internal evaluations indicate a 52.5% decrease in inaccuracies on high-stakes prompts across medicine, law, and finance. The new model also incorporates a "memory sources" feature, allowing users to see and correct the data used for personalized responses, addressing a key concern for enterprise adoption. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Reduces hallucinations in sensitive domains, potentially accelerating enterprise adoption of AI tools like ChatGPT.
RANK_REASON Frontier-lab model release with system card documentation. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
