Auto-translation used

OpenAI O1: A new era of solving complex problems

Today, OpenAI officially unveiled the full version of its newest ChatGPT o1 language model. The model replaces the previous version of o1-preview, launched in September under the code name Strawberry, and is already available to ChatGPT Plus and Team users. Next week, corporate clients and educational institutions will receive access.

ChatGPT o1 is able to break down tasks into simpler steps, analyze different approaches and correct its own mistakes, which makes its problem solving similar to human thinking. Thanks to the new chain of thought learning algorithm, o1 has achieved outstanding success:

  • Mathematics: The model entered the top 500 best students in the United States at the American Mathematical Olympiad (AIME), solving 83% of the problems, which is significantly higher than the GPT-4o (13%).
  • Natural sciences: In GPQA (physics, chemistry, biology) tests, o1 showed accuracy exceeding the level of many candidates and doctors of sciences.
  • Programming: At the International Olympiad in Computer Science (IOI) 2024, the model took 49th place and surpassed 89% of participants on the Codeforces platform.

Compared to the o1-preview version, the new model is more powerful, faster and more accurate. She also gives "reasoned” answers to the images. This version, according to OpenAI, was prepared for greater brevity, which also means a faster response compared to the pre-release version.

In addition, OpenAI has also released ChatGPT Pro with artificial intelligence research intelligence designed mainly for professionals.The company also announced its ChatGPT Pro grant program, which will offer 10 grants to medical researchers at leading institutes. Similar grants are planned for other disciplines.

Comments 7

Login to leave a comment

тағыда басқа салаларға грант бөлінсе, керемет қой

Reply

10 грант керемет

Reply

тағы да бір алға ұмтылыс беретін мүмкіндіктер

Reply

цена конечно очень кусается

Reply

гпт лучший🔥🔥🔥

Reply