OpenAI has officially launched its GPT-4.5 model through a livestream event. However, there’s a catch: users must subscribe to the ChatGPT Pro service for $200 per month to access it. While the model is still in a “research preview” phase, it offers enhanced capabilities in natural conversation, programming, pattern recognition, writing, and problem-solving.
It’s important to note that GPT-4.5 is not designed as a reasoning model like OpenAI’s o1, o3-mini, or the upcoming full version of o3. Despite this, GPT-4.5 is OpenAI’s largest language model yet, making it highly suitable for everyday queries due to its extensive training material.
Credit: OpenAI
According to internal tests, GPT-4.5 exhibits a 24.7% decrease in hallucinations compared to GPT-4.0, with a 34.3% improvement in accuracy. Additionally, approximately 57% of human testers preferred GPT-4.5 over its predecessor.
Credit: OpenAI
In summary, while GPT-4.5 may not be at the forefront of AI technology, it represents an improvement over non-reasoning models. It bridges the gap between more traditional language models and their often restricted counterparts. Although features like AI Voice Mode and video input are currently absent, OpenAI plans to enhance the user experience and potentially introduce these capabilities in future updates.
Considering its early release status, OpenAI is still exploring the limits and applications of GPT-4.5. The model will gradually roll out to all paid ChatGPT plans over the next two weeks and is expected to replace GPT-4.0 for free users once it exits the preview phase.
