🤖 AI Tools· Source: OpenAI Blog

🤖OpenAI Officially Releases o3 Model with Major Reasoning Upgrades

OpenAI's new o3 reasoning model sets new benchmarks in math, coding, and scientific reasoning, initially available to Pro subscribers.

OpenAI o3 Official Release

OpenAI officially launched the o3 model in 2025, a major upgrade to its "thinking model" series.

Key Highlights

  • Math: Scored 96.7% on AIME 2024, far surpassing o1's 83.3%
  • Coding: Achieved a Codeforces rating of 2727, near International Master level
  • Science: Scored 87.7% on GPQA Diamond, first to surpass human expert performance
  • Safety: Dramatically improved through new "deliberate alignment" training

How to Access

o3 is currently available via ChatGPT Pro ($200/month) and the API. A VPN is required to access from mainland China.

Why It Matters

o3's reasoning capability is especially useful for:

  • Complex contract translation and legal document analysis
  • Multilingual customer service automation for cross-border e-commerce
  • Precise understanding and generation of technical documentation