After some quick tests it seems faster than Sonnet 4.5 and slighly less smart than Opus 4.5/4.6.
But given the small 128k context size, I'm tempted to keep using GPT-5.3-Codex which has more than double context size and seems just as smart while costing the same (1x premium request) per prompt.
I have my reservations against OpenAI the company but not enough to sacrifice my productivity.
https://i.imgur.com/mHvtuz8.png
After some quick tests it seems faster than Sonnet 4.5 and slighly less smart than Opus 4.5/4.6.
But given the small 128k context size, I'm tempted to keep using GPT-5.3-Codex which has more than double context size and seems just as smart while costing the same (1x premium request) per prompt.
I have my reservations against OpenAI the company but not enough to sacrifice my productivity.