AI: OpenAI Unveils Its Model O1 For Coding And Reasoning

OpenAI continues to push the limits of artificial intelligence. If yesterday it was generative cinema with Sora that was making waves, today it’s customization on GPT Store that’s drawing interest. But the most notable innovation is the release of its newest model, OpenAI o1. Equipped with unprecedented reasoning abilities, it takes a giant leap in the fields of science and programming. And this is only the beginning.

OpenAI o1: a leap forward

OpenAI did not do things halfway with the launch of its generative cinema with Sora, and why not with its latest AI model, aptly named OpenAI o1. No more hesitations, welcome an AI that takes the time to think before responding.

We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond.

These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. https://t.co/peKzzKX1bu

— OpenAI (@OpenAI) September 12, 2024

The promise? A “chain of thoughts” reasoning that allows this AI to solve complex tasks, particularly in the fields of science, technology, engineering, and mathematics (STEM).

OpenAI also claims that its model excels in high-level tests.

“OpenAI o1 beats GPT-4o in several benchmarks, including doctoral-level scientific subjects in biology, chemistry, and physics“, the company said.

This AI is also a champion of code: it reaches the 89th percentile in programming contests.

Mathematics performance: Correctly solving 83% of problems during an International Mathematical Olympiad test, compared to 13% for GPT-4o.

• 89th percentile in Codeforces programming contests;
• Up to 80% cheaper than the full version with OpenAI o1-mini;
• Security test: score of 84 out of 100 in resistance to “jailbreaking”.

Artificial intelligence and enhanced security

Beyond pure performance, OpenAI also emphasizes security. The OpenAI o1 model is the result of a new training effort, aiming to make it better able to adhere to security and alignment rules. It can thus better apply these rules in complex contexts.

In a resistance test to rule circumvention attempts (jailbreaking), OpenAI o1 scores an impressive 84, compared to 22 for its predecessor, GPT-4o.

This major advance in the field of security has allowed OpenAI to strengthen its collaborations with institutes specializing in artificial intelligence in the United States and the United Kingdom. These agreements aim to assess and improve the security of future AI models before their public release.

A small step for AI, a giant leap for cybersecurity!

OpenAI o1 codes a video game from a prompt. pic.twitter.com/aBEcehP0j8

— OpenAI (@OpenAI) September 12, 2024

AI at the service of developers

The other major asset of OpenAI o1 lies in its practical applications for developers. The AI is not only capable of reasoning, it is also designed for coding.

To meet the specific needs of developers, OpenAI, issuer of WorldCoin (WLD), also offers a faster and more economical version, OpenAI o1-mini, specially optimized for code generation and debugging. This version provides a more accessible solution while offering reasoning capabilities similar to its predecessor.

Already available to ChatGPT Plus subscribers and soon accessible to a wider audience, this model promises to improve the daily lives of developers by automating complex processes. All at a reduced cost of 80% compared to the full version. In short, it boosts productivity without breaking the bank!

In addition, the battle for artificial intelligence is raging, and tech giants are engaged in a fierce struggle. Elon Musk recently launched a challenge to OpenAI and Google, supporting a bill for AI security.