OpenAI, the parent company of ChatGPT, unveils its generative pre-trained transformer OpenAI o1, more faster and powerful reasoning model than the previous versions. 

More efficient 

As per the tweet by OpenAI it is much faster at coding, math & writing along with other features. It has now become more concise in its thinking resulting in a faster response time than the preview launched previously in September 2024. 

“Our testing shows that o1 outperforms o1-preview, reducing major errors on difficult real-world questions by 34 percent,” the tweet read. 

Image upload 

The recently unveiled model supports image upload to facilitate the review and analysis of visual elements. It can read into visuals to describe elements and their nitty-gritty. An example video by OpenAI shows the model analyzing the elements of an image and telling how to make such a thing from scratch. 

OpenAI o1 is now out of preview in ChatGPT.

What’s changed since the preview? A faster, more powerful reasoning model that’s better at coding, math & writing.

o1 now also supports image uploads, allowing it to apply reasoning to visuals for more detailed & useful responses. pic.twitter.com/hrLiID3MhJ

— OpenAI (@OpenAI) December 5, 2024

Further additions 

The company is still working on adding support for tools like web browsing and file upload into OpenAI o1. Additionally, enterprise solutions like Function Calls, Developer Messages, Structured Outputs, and Vision are currently being developed. 

Tests conducted 

Robust tests have been conducted on the model making it rank somewhere in the “medium” risk category. The tests include Disallowed Content, Training Data, Regurgitation, Hallucinations & Bias, the blog by Open AI stated. 

Preparedness 

The model ranks low in cybersecurity and autonomy while medium in Chemical, Biological, Radiological, and Nuclear (CBRN) and persuasion. 

• Cybersecurity and Autonomy: Rated low, indicating limited ability to independently handle or secure sensitive systems.
• CBRN: Moderately effective, showing potential for analyzing structured scenarios but lacking deep specialization.
• Persuasion: Performs at a medium level, capable of generating convincing arguments but not inherently optimized for strategic influence.

These ratings highlight areas where the model could be further developed to enhance utility in specialized or critical applications. 

Availability 

As per the tweet by OpenAI it will be available for use to the Plus and Team users since the launch only while Enterprise and Edu users will have access in a week.