The release of GPT-40 brings several exciting new features. Users will see three models: GPT-3.5, GPT-4, and the latest, GPT-40. The new model is free but has higher rate limits for paying subscribers.
First, let's talk about access. You can use GPT-40 in the ChatGPT interface and the Playground. To compare models, go to Playground, select GPT-3.5, GPT-4 Turbo, and GPT-40 to see differences. Users with paid accounts can make direct comparisons.
GPT-40 is also 50% faster than GPT-4 Turbo. It offers enhanced coding capabilities. Just ask it to code something, and it will generate the code quickly. For example, it can create a Python script or even simple trading bots.
Another feature is generating 3D models. You can ask GPT-40 to make an STL file of an object, like a table. This feature is still in its early stages, so complex objects may not turn out perfectly.
Switching between models mid-conversation is another new feature. If you hit your rate limit with one model, you can switch to another without losing your chat history. This is helpful for long conversations.
Advanced image generation and editing have improved, too. You can make an image and then edit it by selecting parts of the image. For instance, you can add sunglasses to a cartoon cat by selecting the eyes and asking GPT-40 to make the change.
Another cool feature is analyzing emotions in images. While not fully updated, it still offers interesting insights. Data analysis has seen upgrades, too. You can ask GPT-40 to generate synthetic data, analyze it, and produce charts and reports.
Finally, the voice model hasn't shipped yet, but it will soon. The Alpha version will come after May 13th. Desktop and Mac users will get the new voice model in the future. For now, only the text and vision models are available.
In summary, GPT-40 offers faster performance, better coding, image editing, and data analysis. While some features are still in development, the current capabilities show great promise.