Verdict
Built from the measurements below
- Accuracy is too close to call: GPT-5.3-Codex scored 82.5% and GPT-5.4 72.5% on the same 80 questions, a gap within what chance alone could produce.
- They cost about the same per question ($0.01 vs $0.011).
- GPT-5.3-Codex generates text 1.7× faster: 95 tokens per second against 55.
- GPT-5.3-Codex starts answering sooner: 1.4 s to the first token, against 1.7 s for GPT-5.4.
- GPT-5.4 accepts more context: 1.1M tokens against 400k.
Which should you choose?
Choose GPT-5.3-Codex for the lower cost per correct answer ($0.013 against $0.016), faster streaming and a quicker first token. Choose GPT-5.4 for a 1.1M-token context window. The widest subject gap is in medicine: GPT-5.3-Codex answered 7 of 11 correctly, GPT-5.4 4 of 11. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.
Accuracy gap +10.0 points (z = 1.51). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.