OpenAI replaced GPT-5.3 Instant with GPT-5.5 Instant as ChatGPT’s default model on May 5. Hundreds of millions of daily users got a new brain overnight, no opt-in, no banner.
What actually changed
This is a model swap, not a product launch. AIME 2025 math reasoning jumped from 65.4 to 81.2. MMMU-Pro multimodal went 69.2 to 76. The bigger lift is hallucination — OpenAI’s internal tests show 52.5% fewer made-up claims on high-stakes legal, medical, and financial prompts, and a 37.3% drop on real conversations users had previously flagged for factual errors.
Memory got an upgrade too. ChatGPT can now search across past conversations, uploaded files, and connected Gmail to personalize replies. Plus and Pro users see “memory sources” on each answer, and can delete, correct, or skip memory entirely with temporary chats.
API access for developers
Developers hit GPT-5.5 Instant through the chat-latest alias, which auto-points to the current default. The old GPT-5.3 endpoint stays paid-only for three months before retirement. Use cases: high-volume chat backends, RAG pipelines with personal context, multimodal extraction, and agent loops where the old default was burning tokens on weak reasoning.
A non-reasoning default scoring 81 on AIME is the real story — territory that required o-series thinking modes a year ago.
You Might Also Like
- Gpt 5 3 Instant Just Dropped and Chatgpt Finally Stops Being Condescending
- Openai Oauth Turns Your Chatgpt Subscription Into a Free Openai api but Should you use it
- Gpt oss 120b Openai Finally Goes Open Source and its Worth the Wait
- Chatgpt Interactive Visuals Just Dropped Openai Wants 140 Million Weekly Learners to Ditch Static Explanations
- Openai Desktop Superapp Chatgpt Codex and Atlas are Becoming one app
Top AI Product 
Leave a comment