OpenAI Rolls Out GPT-4.5 Model in Phased Launch, Boasts Lower Hallucination Rates

ChatGPT Plus and Team subscribers will gain access to GPT-4.5 next week

by Aishwarya Bhatt · February 28, 2025

All jobs SDE jobs SSE jobs Remote Bengaluru AI jobs

OpenAI has launched a research preview of its latest large language model, GPT-4.5, initially making it available to ChatGPT Pro users and developers worldwide. The Microsoft-backed AI company confirmed the staged rollout on Thursday, revealing plans to expand access to additional user tiers in the coming weeks.

GPT-4.5 Model has numerous novel features such as:

Enhanced Pattern recognition capabilities
Creative insight generation
Emotional intelligence

According to the company's phased deployment strategy, ChatGPT Plus and Team subscribers will gain access to GPT-4.5 next week, while Enterprise and Edu users will receive the update the following week.

CEO Sam Altman addressed the staggered release on X (formerly Twitter), describing GPT-4.5 as "a giant, expensive model" and citing ongoing GPU shortages as the primary constraint.

"We will add tens of thousands of GPUs next week ... this isn't how we want to operate, but it's hard to perfectly predict growth surges that lead to GPU shortages," Altman explained in his post.

A standout improvement in GPT-4.5 is its substantially reduced hallucination rate—the tendency of AI models to generate incorrect or misleading information. OpenAI claims GPT-4.5 achieves a hallucination rate of 37.1%, marking a significant improvement over GPT-4's 61.8% rate and even outperforming the O1 reasoning model's 44% rate.

The latest model supports file and image uploads while enhancing writing and coding assistance capabilities. However, voice and video features—technologies OpenAI has been actively developing—remain absent from this release.

OpenAI's aggressive GPU expansion plans aim to ensure consistent availability as the company extends access to its broader user base. Industry analysts view GPT-4.5 as a crucial intermediate step in OpenAI's development roadmap toward the eventual release of GPT-5, which is expected to push the boundaries of artificial intelligence language modeling and multimodal capabilities further.