If just a couple of days ago GPT-5.3 appeared, today you might think there’s no point in waiting for an update to version 5.4. But here it is already: https://openai.com/index/introducing-gpt-5-4/ — and there’s even a Pro version.
The model has become slightly more expensive, and the context window has increased to a full million tokens, opening up more possibilities for complex tasks.
In addition to standard benchmarks, special attention is given to a new aspect — Computer Use: how the model can interpret screenshots from your PC and give commands to perform various operations. Hopefully, future updates will include improvements to Agent and Atlas; I will definitely test them.
Unfortunately, the model has become even more efficient in token usage within reasoning chains — it now works more effectively. Also, Codex has introduced a /fast mode, which operates 50% faster and uses twice the token limits. Overall, compared to Anthropic, this seems quite impressive.
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby