OpenAI Unveils GPT-4.5: Advanced AI Model Boosts Writing and Knowledge Capabilities
Staff Editor OpenAI Unveils GPT-4.5: A New Milestone in AI Language Models
OpenAI, the leading artificial intelligence research laboratory, has announced the launch of GPT-4.5, its latest and most expansive AI language model to date. The new model is currently available as a research preview for ChatGPT Pro users, with plans for wider distribution in the coming weeks.
Described by OpenAI as their “most knowledgeable model yet,” GPT-4.5 boasts improved writing capabilities and enhanced world knowledge. However, the company has not classified it as a frontier model, indicating that it may not outperform some existing models like o1 or o3-mini in certain aspects.
The new model features a refined personality compared to its predecessors, offering more natural interactions and demonstrating an increased ability to recognize patterns and draw connections. OpenAI positions GPT-4.5 as ideal for writing, programming, and solving practical problems.
In terms of training and development, GPT-4.5 incorporates new supervision techniques alongside traditional methods such as supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). The model also utilizes synthetic data from the o1 reasoning model, codenamed Strawberry.
OpenAI CEO Sam Altman acknowledged GPT-4.5 as a “giant, expensive model” but tempered expectations by stating that it “won’t crush benchmarks.” Despite this, the model has shown improvements in reducing hallucinations compared to GPT-4o and slightly outperforming o1 in this aspect. Human testers have also reported better performance in evaluations compared to GPT-4o.
The rollout plan for GPT-4.5 includes making it available to Plus and Team users next week, followed by Enterprise and Edu users. Additionally, the model will be accessible on Microsoft’s Azure AI Foundry platform.
Looking ahead, OpenAI has ambitious plans for the future, with the launch of GPT-5 anticipated by late May. This upcoming model is expected to integrate the new o3 reasoning model, as the company continues its pursuit of creating more capable AI systems that could potentially be classified as artificial general intelligence (AGI).
As of February 27th, OpenAI updated its GPT-4.5 system card, incorporating additional information from Sam Altman. This launch marks another significant step in the rapidly evolving field of AI language models, promising to push the boundaries of what’s possible in natural language processing and generation.
