GPT-5.4 Pro vs Qwen3 8
GPT-5.4 Pro 
Qwen3 8 Verdict
- GPT-5.4 Pro and Qwen3 8 (max) are level on quality (64.2 vs 63.9).
- GPT-5.4 Pro is stronger in instruction following, knowledge, multimodal, reasoning.
- Qwen3 8 (max) is stronger in agents & tools, coding, composite, human preference, long context.
- Qwen3 8 (max) is 25× cheaper ($2.67 vs $67.50 per 1M blended).
- Qwen3 8 (max) streams 39.4× faster (39 vs 1 tokens per second).
| Metric | GPT-5.4 Pro | Qwen3 8 (max) |
|---|---|---|
| BenchLeader Index | 64.2 | 63.9 |
| Instruction following score | 67.3 | – |
| Knowledge score | 73.2 | 62.3 |
| Multimodal score | 69.8 | 67.2 |
| Reasoning score | 75.1 | 67.7 |
| Agents & tools score | – | 56.1 |
| Coding score | – | 71.7 |
| Composite score | – | 70.8 |
| Human preference score | – | 67.9 |
| Long context score | – | 65.7 |
| Blended price $/M | $67.50 | $2.67 |
| Output speed | 1 tok/s | 39 tok/s |
| Time to first answer | 6.7 s | 53.2 s |
| Context window | 1.1M | 1M |
| Humanity's Last Exam | 44.3% | – |
| SimpleBench | 74.1% | – |
| Epoch Capabilities Index | 158.9 | 156.6 |
| LMArena Text | – | 1480 |
| LMArena Hard Prompts | – | 1502 |
| LMArena Coding | – | 1520 |
| LMArena WebDev | – | 1685 |
| LMArena Vision | – | 1313 |
| LMArena Agent | – | 3.9 |
| LiveBench | – | 78.5% |
| LiveBench Reasoning | – | 88.2% |
| LiveBench Coding | – | 72.9% |
| LiveBench Agentic Coding | – | 64.7% |
| LiveBench Mathematics | – | 91.3% |
| LiveBench Data Analysis | – | 78.4% |
| LiveBench Language | – | 79.7% |
| AA Intelligence Index | – | 40.3 |
| AA-LCR | – | 78.3% |
| MMMU-Pro | – | 82.3% |
| AA-Omniscience | – | 3.4 |
| GPQA Diamond (AA) | – | 92.7% |
| Humanity's Last Exam (AA) | – | 43.0% |
| SciCode (AA) | – | 53.2% |
| LiveCodeBench | – | 87.8% |
| MMLU-Pro | – | 88.6% |
| IOI | – | 73.0% |
| LegalBench | – | 83.6% |
| CorpFin | – | 65.8% |
| TaxEval | – | 75.5% |
| Terminal-Bench 2.1 (Vals) | – | 67.4% |
| SWE-bench (Vals) | – | 85.6% |
| GPQA Diamond (Vals) | – | 93.7% |
| Vals Index | – | 51.8 |
| MultiChallenge | 69.2% | – |
| VISTA | 53.9% | – |
| MultiNRC | 62.3% | – |
| TutorBench | 56.6% | – |
Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.