tech
February 6, 2026
GPT 5.3 Codex vs. Opus 4.6: The Great Convergence
We've tested both models thoroughly—here's our head-to-head Vibe Check
.png)
TL;DR
- OpenAI released GPT-5.3 Codex and Anthropic released Opus 4.6, with both models showing significant improvements and convergence.
- Opus 4.6 combines the precision of Codex with the thoroughness of Opus, while Codex 5.3 has adopted some of Opus's warmth and speed.
- The convergence is driven by the realization that AI behaviors useful for software development are also key for general knowledge work.
- Opus 4.6 has a higher ceiling and variance, making it ideal for maximum upside on hard, open-ended tasks, but can be less reliable.
- Codex 5.3 offers lower variance and higher reliability for steady, autonomous execution, and is notably faster, but may not reach the same creative heights as Opus.
- Internal testing and benchmarks like LFG bench suggest Opus 4.6 performs better on complex, end-to-end builds, while Codex 5.3 is more consistent and reliable.
- Users' preferences often align with their previous experience, with many mixing and matching both models.
Continue reading
the original article