GPT-5.2 vs GPT-5.3-Codex
GPT-5.2 Verdict
- GPT-5.2 and GPT-5.3-Codex are level on quality (61.6 vs 62.5).
- GPT-5.2 is stronger in human preference, reasoning.
- GPT-5.3-Codex is stronger in agents & tools, coding, composite, instruction following, knowledge, long context, multimodal.
- They cost about the same ($4.81 per 1M blended).
- GPT-5.3-Codex streams 1.8× faster (121 vs 68 tokens per second).
| Metric | GPT-5.2 | GPT-5.3-Codex |
|---|---|---|
| BenchLeader Index | 61.6 | 62.5 |
| Agents & tools score | 61.2 | 62.7 |
| Coding score | 53.7 | 54.8 |
| Composite score | 67.5 | 70.1 |
| Human preference score | 67.9 | – |
| Instruction following score | 67.4 | 73.2 |
| Knowledge score | 61.1 | 69.3 |
| Long context score | 67.9 | 68.3 |
| Multimodal score | 59.9 | 63.0 |
| Reasoning score | 61.3 | – |
| Blended price $/M | $4.81 | $4.81 |
| Output speed | 68 tok/s | 121 tok/s |
| Time to first answer | 129.4 s | 58.2 s |
| Context window | 400k | 400k |
| Humanity's Last Exam | 27.8% | – |
| Terminal-Bench | 64.9% | 78.4% |
| SimpleBench | 45.8% | – |
| Remote Labor Index | 2.1% | – |
| WeirdML | – | 79.3% |
| APEX-Agents | 23.0% | 31.8% |
| Epoch Capabilities Index | 153.5 | 156.6 |
| LMArena Text | 1476 | – |
| LMArena Hard Prompts | 1497 | – |
| LMArena Coding | 1515 | – |
| LMArena WebDev | 1417 | 1409 |
| LMArena Vision | 1268 | – |
| AA Intelligence Index | 30.4 | 32.5 |
| IFBench | 75.4% | 75.4% |
| AA-LCR | 82.7% | 83.3% |
| MMMU-Pro | – | 78.5% |
| AA-Omniscience | -0.9 | 10.9 |
| Terminal-Bench Hard | 47.0% | 53.0% |
| GPQA Diamond (AA) | 90.3% | 91.5% |
| Humanity's Last Exam (AA) | 37.7% | 42.5% |
| τ²-Bench Telecom (AA) | 84.8% | 86.0% |
| SWE-Bench Pro | 29.9% | – |
| VISTA | 46.6% | – |
| MultiNRC | 42.2% | – |
| HiL-Bench | – | 4.3% |
| TutorBench | 53.5% | – |
| Kagi LLM Benchmark | 73.3% | – |
| SWE-bench Verified (bash only) | 69.0% | – |
| ARC-AGI-1 | 94.5% | – |
| ARC-AGI-2 | 72.9% | – |
| BFCL Overall | 55.9% | – |
Data as of 2026-09-10. Best configuration of each model; every score links to its source on the model pages.
GPT-5.2 vs GPT-5.3-Codex: questions
- Is GPT-5.2 better than GPT-5.3-Codex?
- GPT-5.2 and GPT-5.3-Codex are level on quality (61.6 vs 62.5). The BenchLeader Index combines every independent quality benchmark; GPT-5.3-Codex is ahead overall as of 2026-09-10, but check the category scores for your use.
- Is GPT-5.2 better than GPT-5.3-Codex for coding?
- GPT-5.3-Codex scores higher in coding (55 vs 54 on the category index, where 50 is average).
- Is GPT-5.2 better than GPT-5.3-Codex for agentic tasks?
- GPT-5.3-Codex scores higher in agentic tasks (63 vs 61 on the category index, where 50 is average).
- Which is cheaper, GPT-5.2 or GPT-5.3-Codex?
- GPT-5.2 is cheaper: $4.81 against $4.81 per million tokens, blended at three input tokens per output token.
- Which is faster, GPT-5.2 or GPT-5.3-Codex?
- GPT-5.3-Codex streams faster: 121 against 68 output tokens per second.
- Which has the larger context window?
- Both accept 400k tokens of context.