All Events
model-release
☆ BENJI

OpenAI o1 Released

Overview OpenAI o1 (codename Strawberry) is a series of models designed for complex reasoning tasks, achieving major breakthroughs through large-scale reinforcement learning and Chain-of-Thought techniques. Key Capabilities Math Olympiad …

2024-09-12

Overview

OpenAI o1 (codename Strawberry) is a series of models designed for complex reasoning tasks, achieving major breakthroughs through large-scale reinforcement learning and Chain-of-Thought techniques.

Key Capabilities

  • Math Olympiad Gold Level: Ranked in top 500 on AIME
  • PhD-level Science Reasoning: Outperforms human experts on GPQA Diamond biology test
  • Chain-of-Thought Reasoning: Internally expands multi-step reasoning process
  • Coding Competition: 89th percentile on Codeforces

Impact

o1 represents a paradigm shift in LLM reasoning — from “memory + pattern matching” to “genuine logical reasoning.”

References