GPT-5.3 Codex vs Step 5 Preview
GPT-5.3 Codex 
Step 5 Preview Verdict
- GPT-5.3 Codex (xhigh) and Step 5 Preview are level on quality (65.4 vs 65.0).
- GPT-5.3 Codex (xhigh) is stronger in agents & tools, coding, instruction following, multimodal.
- Step 5 Preview is stronger in composite, knowledge, long context, reasoning.
- Step 5 Preview is 3.4× cheaper ($1.43 vs $4.81 per 1M blended).
| Metric | GPT-5.3 Codex (xhigh) | Step 5 Preview |
|---|---|---|
| BenchLeader Index | 65.4 | 65.0 |
| Agents & tools score | 79.8 | – |
| Coding score | 61.7 | – |
| Composite score | 70.2 | 84.4 |
| Instruction following score | 73.4 | – |
| Knowledge score | 69.4 | 72.0 |
| Long context score | 68.2 | 70.8 |
| Multimodal score | 63.0 | 60.8 |
| Reasoning score | 74.4 | 82.1 |
| Blended price $/M | $4.81 | $1.43 |
| Output speed | 125 tok/s | – |
| Time to first answer | 50.8 s | – |
| Context window | 400k | 1M |
| WeirdML | 77.9% | – |
| AA Intelligence Index | 32.5 | 43.6 |
| IFBench | 75.4% | – |
| AA-LCR | 83.3% | 88.3% |
| MMMU-Pro | 78.5% | 76.4% |
| AA-Omniscience | 10.9 | 16.4 |
| Terminal-Bench Hard | 53.0% | – |
| GPQA Diamond (AA) | 91.5% | – |
| Humanity's Last Exam (AA) | 42.5% | 46.5% |
| SciCode (AA) | – | 58.9% |
| τ²-Bench Telecom (AA) | 86.0% | – |
| LiveCodeBench | 87.3% | – |
| IOI | 53.8% | – |
| SWE-bench (Vals) | 78.0% | – |
| CritPt | 16.9% | 20.9% |
| GDPval (AA) | – | 53.6% |
| Vibe Code Bench v1.1 | 61.8% | – |
| ALE-Bench | 1655.2 | – |
Data as of 2026-09-19. Best configuration of each model; every score links to its source on the model pages.
GPT-5.3 Codex vs Step 5 Preview: questions
- Is GPT-5.3 Codex better than Step 5 Preview?
- GPT-5.3 Codex (xhigh) and Step 5 Preview are level on quality (65.4 vs 65.0). The BenchLeader Index combines every independent quality benchmark; GPT-5.3 Codex (xhigh) is ahead overall as of 2026-09-19, but check the category scores for your use.
- Which is cheaper, GPT-5.3 Codex or Step 5 Preview?
- Step 5 Preview is cheaper: $1.43 against $4.81 per million tokens, blended at three input tokens per output token.
- Which has the larger context window?
- Step 5 Preview accepts more context: 1M against 400k tokens.