OpenAI rolls out GPT-5.3 Instant to ChatGPT
OpenAI has begun rolling out GPT-5.3 Instant inside ChatGPT, positioning the update as a meaningful step forward in accuracy, responsiveness, and day-to-day usability. The company says the new model delivers 26.8% fewer hallucinations, alongside stronger web reasoning and a more natural conversational tone for users across tiers.
Fewer hallucinations, more direct responses
According to OpenAI, the new release reduces incorrect or fabricated outputs—often referred to as hallucinations—by 26.8%. While the company did not detail the methodology in the announcement, the claim signals continued emphasis on reliability as ChatGPT becomes a default tool for writing, coding, and research workflows.
The update also aims to provide direct answers “without refusals” in more situations, suggesting adjustments to how the assistant handles uncertain queries and sensitive topics. In practice, this could mean fewer dead-end responses and more attempts to be helpful, while still operating within safety guidelines.
Improved web reasoning and tone
GPT-5.3 Instant is described as improving web reasoning, a capability that typically involves synthesizing information, resolving conflicting claims, and following multi-step logic when working with web-based content. OpenAI also highlighted a more natural tone, intended to make conversations feel less robotic and more context-aware.
Availability
OpenAI said the improvements are coming to ChatGPT users broadly, indicating a wide deployment rather than a limited beta. The rollout underscores intensifying competition in consumer AI assistants, where accuracy, speed, and trust are increasingly central differentiators.
Related
