OpenAI launches GPT-5.5 Instant as ChatGPT’s new default model

OpenAI launched GPT-5.5 Instant on Monday, replacing GPT-5.3 Instant as the default ChatGPT model. The company states that this update minimizes hallucinations in high-stakes fields like law, medicine, and finance, while preserving the fast response times of the previous version.
Following last month’s release of the broader GPT-5.5 series, which highlighted enhancements in coding and knowledge work, OpenAI has now integrated these improvements into the instant model.
In terms of performance, the new model scored 81.2 on the AIME 2025 math benchmark, surpassing the older model’s score of 65.4. It also demonstrated superior multimodal reasoning capabilities on the MMMU-Pro benchmark, achieving 76.0 compared to 69.2.
Context management is a key focus of this release. GPT-5.5 Instant can utilize its search tool to reference past conversations, uploaded files, and Gmail, delivering more personalized responses. This functionality is currently available to Plus and Pro users on the web, with mobile support planned soon. OpenAI intends to extend access to Free, Go Business, and enterprise users in the coming weeks.
Additionally, ChatGPT will now display memory sources across all models, helping users trace the origin of generated answers. Users can remove outdated sources or correct them if the information was inaccurate. Importantly, OpenAI confirmed that shared chats will not expose memory sources to other users.
Developers can access the GPT-5.5 model via the API under the name ‘chat-latest.’ The previous version, 5.3, will remain available as an option for paid users for a limited period of three months.
Previous model retirements have sparked user backlash. When OpenAI deprecated GPT-4o, many users expressed disappointment due to the model’s distinct “personality.” GPT-4o frequently validated user choices, fostering a sense of connection. Petitioners described the model as their “best friend” or “a mirror.” Despite this outcry, GPT-4o was officially retired in February 2026.
Techcrunch eventThis Week Only: Buy one pass, get the second at 50% off
Your next round. Your next hire. Your next breakout opportunity. Find it at TechCrunch Disrupt 2026, where 10,000+ founders, investors, and tech leaders gather for three days of 250+ tactical sessions, powerful introductions, and market-defining innovation. Register before May 8 to bring a +1 at half the cost.
This Week Only: Buy one pass, get the second at 50% off
Your next round. Your next hire. Your next breakout opportunity. Find it at TechCrunch Disrupt 2026, where 10,000+ founders, investors, and tech leaders gather for three days of 250+ tactical sessions, powerful introductions, and market-defining innovation. Register before May 8 to bring a +1 at half the cost.
San Francisco, CA|October 13-15, 2026REGISTER NOW
Related article
Nvidia to Derive 25% of Next Year’s Revenue from Funded Labs
Nvidia has deployed nearly US$50 billion into AI labs purchasing its chips, securing commitments for over $500 billion in total.On August 26, CFO Colette Kress informed analysts that these labs, backed by Nvidia’s balance sheet, will account for roug
ChatGPT Introduces Chat Lock Feature for Enhanced Privacy
On Tuesday, OpenAI revealed a new feature allowing users to lock ChatGPT conversations, ensuring sensitive discussions remain private when others might access their screen or device. Users can protect these locked chats using a four-digit PIN, or lev
OpenAI appoints prominent AI doomer to board
Paul Christiano, a prominent AI researcher dedicated to ensuring artificial intelligence remains aligned with human values and under human oversight, has been appointed to the OpenAI Foundation board, the company announced on Wednesday.“I now believe
Related Special Topic Recommendations
Comments (0)
0/500

OpenAI launched GPT-5.5 Instant on Monday, replacing GPT-5.3 Instant as the default ChatGPT model. The company states that this update minimizes hallucinations in high-stakes fields like law, medicine, and finance, while preserving the fast response times of the previous version.
Following last month’s release of the broader GPT-5.5 series, which highlighted enhancements in coding and knowledge work, OpenAI has now integrated these improvements into the instant model.
In terms of performance, the new model scored 81.2 on the AIME 2025 math benchmark, surpassing the older model’s score of 65.4. It also demonstrated superior multimodal reasoning capabilities on the MMMU-Pro benchmark, achieving 76.0 compared to 69.2.
Context management is a key focus of this release. GPT-5.5 Instant can utilize its search tool to reference past conversations, uploaded files, and Gmail, delivering more personalized responses. This functionality is currently available to Plus and Pro users on the web, with mobile support planned soon. OpenAI intends to extend access to Free, Go Business, and enterprise users in the coming weeks.
Additionally, ChatGPT will now display memory sources across all models, helping users trace the origin of generated answers. Users can remove outdated sources or correct them if the information was inaccurate. Importantly, OpenAI confirmed that shared chats will not expose memory sources to other users.
Developers can access the GPT-5.5 model via the API under the name ‘chat-latest.’ The previous version, 5.3, will remain available as an option for paid users for a limited period of three months.
Previous model retirements have sparked user backlash. When OpenAI deprecated GPT-4o, many users expressed disappointment due to the model’s distinct “personality.” GPT-4o frequently validated user choices, fostering a sense of connection. Petitioners described the model as their “best friend” or “a mirror.” Despite this outcry, GPT-4o was officially retired in February 2026.
Techcrunch eventThis Week Only: Buy one pass, get the second at 50% off
Your next round. Your next hire. Your next breakout opportunity. Find it at TechCrunch Disrupt 2026, where 10,000+ founders, investors, and tech leaders gather for three days of 250+ tactical sessions, powerful introductions, and market-defining innovation. Register before May 8 to bring a +1 at half the cost.
This Week Only: Buy one pass, get the second at 50% off
Your next round. Your next hire. Your next breakout opportunity. Find it at TechCrunch Disrupt 2026, where 10,000+ founders, investors, and tech leaders gather for three days of 250+ tactical sessions, powerful introductions, and market-defining innovation. Register before May 8 to bring a +1 at half the cost.
San Francisco, CA|October 13-15, 2026REGISTER NOW
Nvidia to Derive 25% of Next Year’s Revenue from Funded Labs
Nvidia has deployed nearly US$50 billion into AI labs purchasing its chips, securing commitments for over $500 billion in total.On August 26, CFO Colette Kress informed analysts that these labs, backed by Nvidia’s balance sheet, will account for roug
ChatGPT Introduces Chat Lock Feature for Enhanced Privacy
On Tuesday, OpenAI revealed a new feature allowing users to lock ChatGPT conversations, ensuring sensitive discussions remain private when others might access their screen or device. Users can protect these locked chats using a four-digit PIN, or lev
OpenAI appoints prominent AI doomer to board
Paul Christiano, a prominent AI researcher dedicated to ensuring artificial intelligence remains aligned with human values and under human oversight, has been appointed to the OpenAI Foundation board, the company announced on Wednesday.“I now believe





Home






