OpenAI launches GPT-4.5 with enhanced performance, refined personality

GPT-4.5 boasts improved writing capabilities, expanded world knowledge, and a “refined personality,” aiming for more natural and intuitive interactions. OpenAI emphasizes its enhanced ability to recognize patterns and draw connections, making it well-suited for writing, programming, and practical problem-solving.

However, OpenAI has been transparent about GPT-4.5’s limitations. While it offers a 10x improvement in computational efficiency over GPT-4, it does not introduce the breakthrough capabilities expected of a frontier model. Early documentation, later revised, indicated that its performance on preparedness evaluations falls below that of certain internal models and deep research benchmarks.

The model’s development involved “new supervision techniques” combined with traditional methods like supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), similar to those employed for GPT-4o. Notably, GPT-4.5 demonstrates a reduction in hallucinations compared to GPT-4o and a slight improvement over the o1 model.

OpenAI researchers have focused on aligning GPT-4.5 to be a “better collaborator,” emphasizing warmer, more intuitive, and emotionally nuanced conversations. Human evaluations have reportedly shown GPT-4.5 outperforming GPT-4o across various categories.

OpenAI CEO Sam Altman acknowledged that GPT-4.5 is a “giant, expensive model” that “won’t crush benchmarks,” emphasizing its focus on practical improvements.

See also

Following its initial release to Pro users, GPT-4.5 will be rolled out to Plus and Team subscribers next week, followed by Enterprise and Edu users. It is also available on Microsoft’s Azure AI Foundry platform, alongside models from Stability, Cohere, and Microsoft.

This release aligns with OpenAI’s roadmap, which previously outlined plans for a late February launch of GPT-4.5 and a potential late May release of GPT-5. GPT-5, described by Altman as a “system that integrates a lot of our technology,” will incorporate OpenAI’s new o3 reasoning model.

The strategic release of GPT-4.5, with its focus on practical improvements and enhanced collaboration, represents a stepping stone towards OpenAI’s long-term goal of developing artificial general intelligence (AGI). By incrementally combining its various language models and reasoning technologies, OpenAI aims to create a more capable and versatile AI system.