Editorial illustration for OpenAI's GPT-5.3 Instant trims hallucinations 26.8% and reduces refusals
GPT-5.3 Instant Slashes AI Hallucinations by 26.8%
OpenAI's GPT-5.3 Instant trims hallucinations 26.8% and reduces refusals
OpenAI dropped a new model Thursday: GPT-5.3 Instant. The headline figure is a 26.8% cut in AI "hallucinations," those maddening factual fabrications. Just as critical, the update tackles something far squishier—a patronizing "cringe" tone, adjusted using direct user complaints. The company also told its model to stop refusing questions so often.
User feedback showed a 22.5% decrease in hallucinations when answering queries using web search.
That 26.8% hallucination reduction is a concrete engineering win. But OpenAI’s blog post admits its own benchmarks miss the daily friction users report—the tone, the refusals, the flow. The real shift here is prioritizing how the AI communicates, not just what it knows. Fixing how it "feels" could matter more for regular use than any single metric.
Related Reading
Browse all OpenAI news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic OpenAI hires 630 ex-Meta staff as ChatGPT memory may turn data into ads
Business & Startups
Related Topic Sen. Warren urges Trump admin to expand AI credit after OpenAI letter
Business & Startups
From Archives Google launches Gemini 3.1 Flash Lite, priced at one‑eighth of Gemini 3.1 Pro
From Archives Pixel 10 adds Circle to Search and Gemini agentic tools for grocery orders
From Archives OpenAI's AI data agent, built by two engineers, now used daily by 4,000 staff
Business & Startups
From Archives Joe Gebbia seen with metallic device as OpenAI, Jony Ive partnership looms
Business & Startups
Common Questions Answered
How much have hallucinations been reduced in GPT-5.3 Instant?
OpenAI reports that GPT-5.3 Instant has trimmed hallucinations by 26.8%. This reduction aims to improve the model's reliability and accuracy in generating responses.
What specific improvements does GPT-5.3 Instant bring to the ChatGPT experience?
The update focuses on improving tone, relevance, and conversational flow, moving away from what OpenAI calls a 'cringe' tone. These nuanced changes are designed to make ChatGPT feel more helpful and natural in everyday interactions.
Which versions of the model receive the GPT-5.3 Instant upgrade?
Currently, only the Instant model receives the 5.3 upgrade, while other variants remain unchanged. Users will experience these improvements on both ChatGPT and the API platform.
Further Reading
- Papers with Code - Latest NLP Research — Papers with Code
- Hugging Face Daily Papers — Hugging Face
- ArXiv CS.CL (Computation and Language) — ArXiv
