OpenAI released OpenAI GPT-5.3 Instant on March 4, 2026, as an upgrade to the most widely used ChatGPT model. The update aims to reduce hallucinations, limit unnecessary refusals, and improve writing and conversational clarity for users worldwide.
OpenAI AI update 2026 introduces GPT-5.3 Instant
OpenAI introduced the latest update for ChatGPT’s default model. The company said the upgrade focuses on improving everyday conversations and response quality.
Users can access the model immediately in ChatGPT. Developers can also use the system through the API under the name “gpt-5.3-chat-latest.”
The company confirmed that GPT-5.2 Instant will remain available for paid users in the legacy model section. OpenAI plans to retire that version on June 3.
How OpenAI GPT-5.3 Instant improves conversations
The new model adjusts how ChatGPT responds during discussions. OpenAI said earlier versions sometimes produced defensive replies or unnecessary refusals.
Now OpenAI GPT-5.3 Instant aims to answer questions directly when safety guidelines allow a response. The system also reduces long warnings and excessive disclaimers that interrupt conversation flow.
OpenAI stated that the model now generates more contextual results during web searches. As a result, users receive clearer answers and fewer conversational dead ends.
AI update improves accuracy and reduces Hallucinations
OpenAI also reported improvements in factual accuracy. Internal tests showed that the model reduced hallucination rates by 26.8% when web search was enabled.
When web search was disabled, hallucinations declined by about 19.7% compared with earlier models. Another evaluation based on user-flagged conversations showed hallucinations dropped by 22.5% with web access and 9.6% without it.
These tests included knowledge-heavy fields such as medicine, law, and finance. OpenAI said the improvements help deliver more reliable responses in factual discussions.
Writing capabilities expanded in OpenAI GPT-5.3 Instant
OpenAI said the model also improves writing tasks. The system performs better in drafting stories, editing documents, and producing longer structured responses.
According to the company, the system can switch between analytical tasks and creative writing while maintaining clarity. The update also aims to reduce overly assertive tones that appeared in earlier versions.
However, OpenAI acknowledged that language fluency still varies across regions. Responses in some languages, including Japanese and Korean, may still sound literal while improvements continue.
