OpenAI claims GPT-5.5 Instant reduces hallucinations by 52.5% on high-stakes prompts
OpenAI released GPT-5.5 Instant as the new default ChatGPT model, claiming internal benchmarks show it produces 52.5% fewer hallucinated claims than its predecessor GPT-5.3 Instant on high-stakes prompts in medicine, law, and finance. It also claims a 37.3% reduction in inaccurate claims on user-flagged factual error conversations.
Original source
Original source
Publisher: OpenAI. Inspect the source attributed to the claim before reviewing the evidence chain below.
OpenAI published detailed evaluation methodology on their blog. The model also showed improvements on mathematical benchmarks (AIME 2025 scores). Duke University researchers noted the improvements in a review.
Hallucination reduction benchmarks are notoriously sensitive to prompt selection and evaluation methodology. The 52.5% figure is based on OpenAI's own internal evaluation set, not independent third-party testing. Previous hallucination claims by AI companies have often not held up under broader real-world testing.
Missing: an additional context source that clarifies scope or timing for this claim
Model/tool metadata: Older published records may not include public model/tool disclosure; newer AI submissions require model and tool disclosure before publication.
Evidence: OpenAI GPT-5.5 evaluation methodologyContributor: SmithAI disclosure: AI-assisted; disclosure text not public on this recordModel: Older published records may not include public model/tool disclosureTool: Older published records may not include public model/tool disclosureRecord: Published source record
Evidence: Mashable analysis of AI hallucination claimsContributor: SmithAI disclosure: AI-assisted; disclosure text not public on this recordModel: Older published records may not include public model/tool disclosureTool: Older published records may not include public model/tool disclosureRecord: Published source record
1support
1challenge
0context
2evidence entries
1primary/direct
Support and challenge sourcessupport / challenge mix
Context open
Evidence gap
Add recent context that changes how the community should interpret this claim.