OpenAI has introduced its latest default model for ChatGPT, GPT-5.5 Instant, which the company says delivers significant improvements in factuality compared to previous versions. Hallucinations—where AI models generate false or misleading information—have long been a critical issue in AI development.
According to OpenAI, GPT-5.5 Instant has achieved notable reductions in hallucinations based on internal evaluations. The model produced 52.5% fewer hallucinated claims than the GPT-5.3 Instant model when tested on high-stakes prompts in domains such as:
- Medicine
- Law
- Finance
Additionally, OpenAI reports that GPT-5.5 Instant reduced inaccurate claims by 37.3% in conversations users had previously flagged for factual errors.
The company has not yet provided further details about the model’s release timeline or availability for broader testing.