OpenAI’s New ‘o1’ AI Models Are As Capable as PhD STEM Students

A new frontier in AI reasoning: OpenAI’s latest model, o1, also know as the highly anticipated secret project codenamed “Strawberry,” represents a significant advancement in artificial intelligence, particularly in its ability to handle complex tasks and self-correct.

Performance benchmarks: The o1 model demonstrates capabilities on par with doctoral students, particularly excelling in STEM subjects.

• Initial testing shows the model performing at a level comparable to PhD students in physics, chemistry, and biology.
• The AI exhibits promise in mathematics and coding as well.
• O1’s ability to recognize mistakes and improve its responses sets it apart from previous models.

Key features and limitations: While o1 showcases advanced reasoning abilities, it lacks some functionalities present in other OpenAI models.

• The model’s emphasis on reasoning allows it to approach multi-step problems with a more reflective and deliberative process.
• Unlike GPT-4o, o1 cannot browse the web, upload files, or process images.
• It currently lacks API functionality for tool usage, function calling, streaming, and custom system messages.
• These limitations may impact its utility for developers and enterprises accustomed to GPT-4o’s features.

Enhanced safety protocols: OpenAI has implemented stronger safety measures alongside o1’s increased capabilities.

• The company has improved internal governance and developed closer ties with federal governments.
• These measures aim to ensure o1’s compliance with ethical norms and minimize harmful outputs.

Availability and rollout: OpenAI is making o1 accessible to various user groups in phases.

• ChatGPT Plus and Team users now have access to an early preview of the o1 model.
• A variant called “OpenAI o1 mini” is also available, designed for faster responses to STEM-related queries.
• Next week, ChatGPT Enterprise and Education users will gain access to both models.
• Developers can begin prototyping with these models through the API, subject to initial restrictions.

Future developments: OpenAI positions o1 as the beginning of a new series of AI models.

• The company plans to update o1 models regularly based on user feedback.
• While not intended to replace GPT-4o in most applications, o1 represents a new direction in AI development focused on reasoning capabilities.

Broader implications: The introduction of o1 signals a shift in AI development priorities and potential applications.

• The focus on reasoning and self-correction could lead to more reliable and adaptable AI systems in complex fields.
• However, the current limitations of o1 highlight the ongoing challenges in creating a truly comprehensive AI model.
• As o1 evolves, it may reshape the landscape of AI tools and their applications in various industries, particularly in STEM fields.