How much do hallucinations decrease with OpenAI's GPT-5.3 Instant?

GPT-5.3 Instant reduces hallucinations by up to 26.8% based on OpenAI's internal evaluations.

What are the key improvements in OpenAI's GPT-5.3 Instant?

The GPT-5.3 Instant model improves accuracy, reduces hallucinations, enhances tone, relevance, and conversational flow, while also decreasing refusals.

Is OpenAI focusing more on accuracy than performance with GPT-5.3 Instant?

Yes, OpenAI is prioritizing accuracy and conversational reliability over raw performance gains with the release of GPT-5.3 Instant.

Home / Technology / GPT-5.3 Instant Cuts Hallucinations by 26.8%

GPT-5.3 Instant Cuts Hallucinations by 26.8%

4 Mar

•

Summary

GPT-5.3 Instant reduces hallucinations by up to 26.8% and improves reliability.
The new model enhances tone, relevance, and conversation with fewer refusals.
OpenAI prioritizes accuracy and conversational reliability in its latest AI update.

GPT-5.3 Instant Cuts Hallucinations by 26.8%

OpenAI has launched GPT-5.3 Instant, its most utilized model, which demonstrates a notable reduction in hallucinations by up to 26.8% compared to its predecessor. This advancement underscores OpenAI's strategic shift towards prioritizing accuracy and conversational dependability over mere performance enhancements.

The GPT-5.3 Instant model, the default for most ChatGPT users, also boasts improvements in tone, relevance, and conversational flow, while decreasing instances of model refusals. Internal evaluations, including high-stakes domains like medicine and finance, as well as user feedback, confirm these improvements. Reliability is boosted by a more effective balance between internet information and the model's internal reasoning.

This update significantly refines the user experience by addressing aspects like tone and conversational consistency. OpenAI aims to move away from an overly cautious or preachy tone, ensuring more direct and helpful responses. While the model shows limitations in certain languages like Korean and Japanese, OpenAI is actively working on broader updates for other ChatGPT models.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.