GPT-5.3 INSTANT CUTS HALLUCINATIONS BY 26.8% AS OPENAI PRIORITIZES ACCURACY OVER SPEED

OpenAl has rolled out GPT-5.3 Instant, positioning it as a meaningful refinement of its most widely deployed model rather than a raw performance leap.

- 

According to the company, the updated system reduces hallucinations by up to 26.8 percent compared to its predecessor, while also cutting down on unnecessary refusals and awkward, off-tone responses that users have increasingly criticized in recent months.

GPT-5.3 Instant effectively serves as the default model for many ChatGPT users and remains one of the most heavily utilized models through OpenAl's API. Instead of focusing on headline-grabbing benchmark jumps, the release centers on reliability, conversational stability, and response accuracy — a strategic pivot that reflects broader industry pressure to make Al systems more dependable in real-world use.

SHIFTING FROM RAW PERFORMANCE TO RESPONSE QUALITY

Earlier model iterations often emphasized speed, scale, and expanded capability. GPT-5.3 Instant signals a recalibration. OpenAl says the model has been tuned to deliver more grounded responses, reducing the frequency of fabricated facts, incorrect citations, and confidently stated inaccuracies — commonly referred to as hallucinations.