Could anyone guide me on the best alternative among the currently available gpt models to replace GPT-4 Vision Preview, which has now been deprecated ?
Have you looked at the public documentation?
1 Like
Welcome to the developer community @navaal
Almost all the gpt-4 models that came after the vision preview have vision capabilities.
The latest ones are gpt-4o and gpt-4o-mini.
Iād recommend checking the docs link shared by @merefield above
1 Like
gpt-4-turbo-2024-04-09 has vision capability (without vision in the name). It is built on the same gpt-4-turbo platform as gpt-4-1106-vision-preview. It has the same $10-$30/1M pricing as gpt-4-vision-preview, reflecting its computational performance.
Still, gpt-4o has different vision abilities that you may find advantageous in particular application, at less expense.
gpt-4o-mini has higher image cost than the latest gpt-4o, due to peculiarity about the competitive pricing policies OpenAI has placed for image analysis applications.
The best alternative for you may be the one with the least changes in the model architecture, if you want to continue your application with similar performance, without needing complete re-evaluation and redesign of prompting to reach objectives.