New Default Model
OpenAI has officially transitioned to GPT-5.5 Instant as the primary model powering ChatGPT, stepping in to replace the prior GPT-5.3 Instant. This strategic
shift is driven by a commitment to refine the AI's capabilities, particularly in mitigating instances of 'hallucination' – where the AI generates incorrect or nonsensical information – across critical domains like law, medicine, and finance. Crucially, this enhanced accuracy has been achieved without sacrificing the responsiveness and speed users expect. The release of GPT-5.5 Instant follows the broader rollout of the GPT-5.5 series last month, which already introduced significant advancements in areas such as complex code generation, sophisticated reasoning, and knowledge-intensive tasks, setting the stage for this even more refined iteration.
Performance Boost
The performance gains offered by GPT-5.5 Instant are substantial, as evidenced by its impressive scores on key industry benchmarks. On the AIME 2025 mathematics benchmark, the new model achieved a remarkable 81.2 points, a significant leap from the 65.4 points recorded by its predecessor. Furthermore, its proficiency in multimodal reasoning has also seen a notable improvement, demonstrated by a score of 76 on the MMMU-Pro benchmark, compared to the previous 69.2. These metrics strongly indicate that GPT-5.5 Instant is exceptionally well-equipped to handle highly complex queries that require intricate calculations, logical deduction, and the interpretation of diverse multimedia inputs, offering a more robust and capable AI experience.
Contextual Intelligence
A particularly groundbreaking advancement in GPT-5.5 Instant lies in its enhanced contextual awareness. This new capability allows the AI to leverage integrated search functionalities, enabling it to draw information from a user's previous conversations, any uploaded documents, and even data from Gmail. This sophisticated integration permits ChatGPT to deliver responses that are not only accurate but also highly personalized and contextually relevant to the individual user's needs and history. Initially, this advanced feature will be available to Plus and Pro subscribers via the web interface, with a mobile app update anticipated shortly thereafter. The company also plans to extend this personalized contextual functionality to Free, Go Business, and enterprise users in the coming weeks.
Enhanced Transparency
In a significant move towards user empowerment and trust, ChatGPT is now incorporating 'memory source information' for all its models, including GPT-5.5 Instant. This new feature provides users with clear insights into the origins of the AI's responses, allowing them to understand precisely where the information is being drawn from. More importantly, users will have the ability to review, delete, or correct any outdated or inaccurate information held within the AI's memory. It's important to note that these memory details remain strictly private to the user and are not exposed when conversations are shared with others, ensuring user privacy remains paramount.
Developer Access
For those building applications and services, GPT-5.5 Instant is readily accessible through an API under the designation 'chat-latest'. This ensures that developers can harness the power of OpenAI's most advanced conversational AI. To facilitate a smooth transition for existing integrations and workflows, the previous GPT-5.3 model will remain available to paid users for a limited period of three months. This phased approach allows developers ample time to adapt their systems to the new API and benefit from the enhanced capabilities offered by GPT-5.5 Instant.
