San Francisco – OpenAI announced on Wednesday a major upgrade to its flagship product, ChatGPT, introducing the GPT-5.5 Instant model as the new default for all users. This upgrade aims to tackle long-standing issues of “hallucination” and verbosity. Official data reveals that factual inaccuracies have been reduced by 52.5% compared to previous iterations, with significant performance leaps in high-stakes domains such as medicine, law, and finance.
Leap in Reasoning Logic and Output Efficiency
The new model demonstrates a substantial enhancement in problem-solving through “error tracing and correction.” In complex algebraic tasks, GPT-5.5 Instant now proactively reviews user steps to pinpoint errors rather than simply stating a lack of solution. Furthermore, output efficiency has been optimized, using 30.2% fewer words while maintaining informational integrity. The interaction style has also become more direct, reducing gratuitous emojis and proactive follow-up questions.
Smart Routing and “Memory Sources” Features
ChatGPT now features a “Smart Routing” mechanism that automatically switches to the GPT-5.5 Thinking model for deep reasoning on complex queries without consuming paid quotas. Additionally, the new “Memory Sources” characteristic allows users to see which historical conversations influenced a specific response. A “Sources” button enables users to directly delete or modify outdated memories, ensuring a more personalized and accurate experience while protecting privacy.
Advanced Safety Ratings and Access Tiers
GPT-5.5 Instant is the first Instant-tier model to be rated as “High Capability” in cybersecurity and biology, reflecting its advanced assistive capabilities. The model is now available to all users: Free users receive 10 messages every 5 hours, while Plus subscribers get 160 messages every 3 hours. For “Pro” users, OpenAI offers unlimited access with a 128K context window, while the powerful Thinking model remains an option for professional and business tiers with expanded reasoning capabilities.
