From: schmidtisblog.de
Published in cooperation with schmidtisblog.de
OpenAI has released a new ChatGPT model that is intended to reduce one problem in particular with AI: incorrect or made-up answers. GPT-5.3 Instant integrates web information more closely into its responses and is said to work more reliably.
OpenAI has introduced significant improvements to everyday AI with the release of GPT-5.3 Instant. According to the company, the new standard ChatGPT model and API reduced the frequency of hallucinations by up to 26.8 percent. Instead of providing simple lists of links, AI now integrates web content more deeply into its reasoning. Reliability has increased noticeably, particularly in sensitive areas such as financial or medical topics. With these adjustments, OpenAI is responding directly to user criticism, which previously often complained about an inaccurate or incorrect factual basis.
The didactic introductions and overly cautious rejections are a thing of the past. GPT-5.3 Instant dispenses with moral preamble and provides much more precise answers. Previously, the model often refused harmless requests for no apparent reason or seemed unnatural due to his didactic tone. These disruptive behaviors have been reduced to allow for more fluid dialogue. The answers are overall shorter and avoid unnecessary repetitions that often hindered the flow of the conversation in the past.
The new version replaces the previous model GPT-5.2 Instant. Access via the API is possible via the gpt-5.3-chat-latest endpoint. For users who want to continue working with the old model, it will remain available under the legacy options for a transition period of three months. However, as of June 3, 2026, support for GPT-5.2 will be completely discontinued. While the Instant version is being made available worldwide, the updates for the Thinking and Pro models are still pending but are expected to follow soon.
Interestingly, the current safety report shows slight deteriorations in certain areas such as sexual content or suicide prevention compared to the previous version. However, OpenAI emphasizes that these statistical fluctuations will still be optimized after the full rollout. The company’s current focus is more on content robustness than on pure performance metrics. At the same time, OpenAI is already hinting at the release of GPT-5.4 via social media channels, indicating an exceptionally short development cycle this year.
About the Author
