OpenAI launches two mini AI models; Know their benefits

OpenAI unveiled its two new Artificial Intelligence (AI) models, o3 and o4-Mini, on Wednesday. Both of the new models are OpenAI's most sophisticated as well as reasoning-centric models, which have the power of visual reasoning, ie, now the new models can answer complex questions precisely even by viewing images. Both new models are the updated versions of older O1 as well as O3-Mini models, and currently, this service will be available for ChatGPT paid subscribers, Plus, Pro, as well as Team users.

Visual Reasoning implies that from now on, not just text, but information embedded in the image will be understandable for AI models as well. OpenAI claims that o3 and o4-Mini models have reached a level of capability for the first time that enables them to leverage all tools available within ChatGP,T like web search, executing Python code, image analysis, file interpretation, as well as image generation.

These models can now also do many difficult things, such as reading a notebook placed upside down, reading text on a board visible from a blur or distance, identifying a specific question from a large list, extracting a timetable from a photo of a bus, solving a puzzle.

OpenAI has claimed that the o3 and o4-Mini models have performed better on many benchmarks such as MMMU, MathVista, CharXiv compared to GPT-4o and the old o1 model, although the company has also admitted that sometimes these models can unnecessarily edit the image too much. Can get entangled in long-term thinking and sometimes can also give wrong answers by misunderstanding the image.