ML//code generation//Codex

Codex is OpenAI's GPT-3 fine-tuned on GitHub code (2021), the model that powered the original GitHub Copilot. It introduced HumanEval, on which its 12B parameters solved 28.8% of the problems, proving that code LLMs were viable.

Codex is OpenAI's GPT-3 fine-tuned on GitHub code (2021), the model that powered the original GitHub Copilot. It introduced HumanEval, on which its 12B parameters solved 28.8% of the problems, proving that code LLMs were viable.

It is now obsolete (GPT-4 and Claude are far better) but historically critical, because it started the code AI industry.