OpenAI’s reasoning fashions, o1 and o3-mini, are useful instruments for individuals who depend on ChatGPT for extra advanced prompts, together with coding, math, and even multi-step textual content prompts, resembling working by an issue or constructing an itinerary. To make the expertise extra useful, OpenAI has deployed small however mighty updates.
OpenAI introduced through an X put up that OpenAI o1 and o3-mini now assist file and picture uploads in ChatGPT. The power to add pictures is likely one of the strongest ChatGPT options as a result of it permits the chatbot to help as you’re employed with on a regular basis assets, whether or not a photograph, doc, graph, or PDF:
Two updates you will like—
📁 OpenAI o1 and o3-mini now assist each file & picture uploads in ChatGPT
⬆️ We raised o3-mini-high limits by 7x for Plus customers to as much as 50 per day— OpenAI (@OpenAI) February 12, 2025
Combining the power to add information and the superior reasoning of ChatGPT opens a brand new vary of use instances. For instance, when importing a protracted analysis paper, ChatGPT can present a extra thorough and significant evaluation than earlier than. One other use case is a posh math drawback you are attempting to unravel on paper. Now, you may add a photograph and get correct help.
The perfect half is that, not like when most new options are first restricted to paid subscribers, free customers can entry the o3-mini mannequin with uploads.
As a free person, click on “Cause” within the message textbox and fasten a photograph or doc to your immediate. Paid customers can choose the mannequin from the mannequin picker and add a picture as they’d recurrently.
Within the announcement, OpenAI additionally raised the utilization limits for o3-mini-high, making it seven instances better for Plus customers with a capability of as much as 50 messages per day.
The o3-mini mannequin with high-reasoning effort carried out considerably higher than o3-mini within the benchmarks OpenAI posted when it first launched the mannequin, coming near o1’s efficiency, generally even surpassing it, as seen within the AIME 2024.
