OpenAI Unveils o1 — An AI Model Capable of Thinking

AI startup OpenAI has introduced a new large language model, o1, trained using reinforcement learning to perform complex reasoning.

The neural network is capable of “thinking” — it can create long internal chains of thought while analyzing a question, the company claims.

“o1 avoids some ‘reasoning traps’ as it can effectively self-check, spending more time analyzing parts of the question. The model approaches problem-solving holistically, planning and executing a series of actions over extended periods of time.”

The new tool from OpenAI is suitable for tasks that involve multiple sub-tasks, such as detecting privileged emails or brainstorming marketing strategies.

“The thought process of o1 takes seconds, but we aim for future versions to think for hours, days, or even weeks. The cost of reasoning will be higher, but what price are you willing to pay for a new cancer drug? For revolutionary batteries? For a proof of the Riemann Hypothesis? AI can be more than just chatbots,” said Noam Brown, a researcher at OpenAI.

According to OpenAI, during the International Math Olympiad selection exam, o1 solved 83% of the problems, while GPT-4o solved only 13%. In competitive programming on the Codeforces platform, it outperformed 89% of participants.

The neural network combines a family of models. The o1-preview early version was released on September 12, with additional versions planned for the future.

Currently, the model has some limitations: it cannot analyze web pages, files, or images.

The new tool is more expensive than its predecessor: the o1-preview API costs $15 per million input tokens (three times more than GPT-4o) and $60 per million output tokens (four times more than GPT-4o). It is available to ChatGPT corporate users through Plus or Team.

September 13, 2024 | 11:09 Update:

Following the release of the new AI model by Sam Altman‘s company, the token for his other project Worldcoin (WLD) saw a 16% increase in value, though the price slightly retreated afterward.

The model was previously code-named Strawberry. Before that, the project was known as Q* and was considered a breakthrough for creating general artificial intelligence that “could pose a threat to humanity.”

It was also revealed that OpenAI is negotiating to raise $6.5 billion at a $150 billion valuation.

It’s worth recalling that Sam Altman, the CEO of the startup, plans to implement a massive project for developing AI hardware and systems worth tens of billions of dollars.