codex-1
OpenAI开发的编程专用模型，是经过强化学习优化的o3变体，作为OpenAI Codex产品的底层模型
Core Facts
Timeline (last 90 days)
codex-1 introduced RLEF (Reinforcement Learning from Execution Feedback), where code generated by the model is actually executed and the test results are fed back to the model as reward signals.
codex-1 significantly outperforms general-purpose models optimized only for conversation on software engineering benchmarks such as SWE-bench.
codex-1 used real code repositories, test case execution feedback, and code review data during its training process.
All Facts (4)
The OpenAI Codex desktop version is based on the codex-1 model, which has been fine-tuned with reinforcement learning and specifically optimized for code generation, debugging, and engineering tasks.
80%Unverifiedcodex-1 introduced RLEF (Reinforcement Learning from Execution Feedback), where code generated by the model is actually executed and the test results are fed back to the model as reward signals.
50%Unverifiedcodex-1 significantly outperforms general-purpose models optimized only for conversation on software engineering benchmarks such as SWE-bench.
50%Unverifiedcodex-1 used real code repositories, test case execution feedback, and code review data during its training process.
50%