OpenAI officially unveiled the GPT-4.5 model during a live broadcast a few hours ago. This marks the final non-chain-of-thought model from OpenAI, as all future iterations starting from GPT-5 will belong to the chain-of-thought category, endowed with advanced reasoning capabilities.
GPT-4.5 serves as a transitional model preceding the release of GPT-5. According to OpenAI, early testing indicates that interactions with GPT-4.5 feel more natural, its knowledge base is more extensive, and its alignment with user intent is remarkably high. Additionally, it exhibits a heightened sense of emotional nuance, making it particularly well-suited for tasks such as writing, programming, and problem-solving, while also significantly reducing hallucinations.
Although GPT-4.5 is not a dedicated reasoning model, OpenAI engineers note that it still possesses a wealth of advanced knowledge that can be leveraged for inferential tasks. However, unlike the “o” series models, which are optimized for scientific and mathematical applications, GPT-4.5 is designed as a versatile general-purpose model.
Regarding multimodal capabilities, GPT-4.5 supports internet access for retrieving real-time information, file and image uploads, and programming-related tasks within the canvas platform. However, at present, it does not support voice mode, video comprehension, or screen sharing.
GPT-4.5 is ready!
good news: it is the first model that feels like talking to a thoughtful person to me. i have had several moments where i've sat back in my chair and been astonished at getting actually good advice from an AI.
bad news: it is a giant, expensive model. we…
— Sam Altman (@sama) February 27, 2025
Sam Altman remarked that GPT-4.5 is an expansive and computationally expensive model. OpenAI is eager to make it available to both Plus and Pro subscribers, but due to the increased demand, GPU resources are currently insufficient. To address this, OpenAI plans to add tens of thousands of GPUs next week, after which GPT-4.5 will be rolled out to ChatGPT Plus subscribers.
Unlike chain-of-thought models that derive answers through step-by-step reasoning, GPT-4.5 relies on a technique known as unsupervised learning, which entails expanded training datasets and increased computational requirements.
Nonetheless, the reduction of AI hallucinations is undoubtedly a welcome improvement. Hallucinations refer to instances where the model fabricates nonexistent information or provides responses entirely unrelated to user queries. The enhanced reliability of GPT-4.5 in mitigating such occurrences is a significant step forward.
Currently, ChatGPT Pro subscribers can access GPT-4.5 across web, mobile, and PC platforms. Support for ChatGPT Plus and Team subscriptions is expected next week, followed by Enterprise and Education editions. Free-tier users, however, may need to wait a little longer.
Related Posts:
- Cybersecurity Alert: Python Libraries Exploited for Malicious Intent
- OpenAI announces a bug bounty program, providing a bug bounty ranging from $200 to $20,000
