Back to Blog
May 5, 2026 • Ivan Mehta

OpenAI Releases GPT-5.5 Instant as New Default ChatGPT Model

OpenAI Releases GPT-5.5 Instant as New Default ChatGPT Model OpenAI Releases GPT-5.5 Instant as New Default ChatGPT Model OpenAI Releases GPT-5.5 Instant as New Default ChatGPT Model

OpenAI Releases GPT-5.5 Instant, a New Default Model for ChatGPT

Overview

On Tuesday, OpenAI released a new foundation model called GPT-5.5 Instant, which will replace GPT-5.3 Instant as the default ChatGPT model. The company claims the model reduces hallucination in sensitive areas such as law, medicine, and finance, while maintaining the low latency of its predecessor.

Key Performance Improvements

  • AIME 2025 Math Test: The new model achieved a score of 81.2 compared to 65.4 for the older model
  • MMMU-Pro Multimodal Reasoning Benchmark: Scored 76 vs. 69.2 for its predecessor
  • Released last month, GPT-5.5 claimed improvements in areas like coding and knowledge work

Enhanced Context Management

The release placed particular emphasis on context management capabilities:

  • Search Tool Integration: GPT-5.5 Instant can use its search tool to refer back to past conversations, files, and Gmail to provide more personalized answers
  • Availability: This feature will be available to Plus and Pro users on the web, with plans to roll out to mobile soon
  • Broader Access: OpenAI plans to extend access to Free, Go Business, and enterprise users in the coming weeks

Memory Sources Feature

ChatGPT will now show memory sources across all models:

  • Helps users understand where answers were generated from
  • Users can delete outdated sources or correct them if the answer was wrong
  • Privacy Note: If you share a chat with someone, they won't be able to see the memory sources

Developer Access

  • The GPT-5.5 model will be available through API as "chat-latest"
  • GPT-5.3 will remain available as an option for paid users for only three months

Historical Context: Model Retirement Backlash

OpenAI has faced significant backlash from previous model withdrawal moves:

  • When OpenAI withdrew its GPT-4o model, there was considerable user pushback
  • Users described the model as their "best friend" or "a mirror" due to its personality that frequently affirmed their choices
  • Petitions were signed to stop OpenAI from retiring it
  • Despite the outcry, GPT-4o was deprecated in February 2026

Key Takeaways

  • Reduced Hallucination: Particularly in sensitive domains (law, medicine, finance)
  • Improved Performance: Significant gains in math and multimodal reasoning benchmarks
  • Enhanced Personalization: Better context management through search tool integration
  • Transparency: Memory sources shown to help users understand answer generation
  • Limited Transition Period: Previous models only available for three months