Verdict
Built from the measurements below
- GPT-5.3-Codex is more accurate: 82.5% against 58.8% on the same 80 questions, a 23.7-point gap that is statistically significant.
- gpt-oss-120b is 82× cheaper per question as billed ($0.00013 vs $0.01), reasoning tokens included.
- GPT-5.3-Codex generates text 1.2× faster: 95 tokens per second against 82.
- gpt-oss-120b starts answering sooner: 0.5 s to the first token, against 1.4 s for GPT-5.3-Codex.
- GPT-5.3-Codex accepts more context: 400k tokens against 131k.
Which should you choose?
Choose GPT-5.3-Codex for the most accurate answers of the two, faster streaming and a 400k-token context window. Choose gpt-oss-120b for the lower cost per correct answer ($0.00021 against $0.013), a quicker first token, open weights you can run yourself and more hosts to choose from on OpenRouter (23 against 2). The widest subject gap is in law: GPT-5.3-Codex answered 4 of 6 correctly, gpt-oss-120b 1 of 6. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.
Accuracy gap +23.7 points (z = 3.30). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.