GPT-5.4 vs Qwen3.8 Max
GPT-5.4 
Qwen3.8 Max Verdict
- GPT-5.4 (xhigh) and Qwen3.8 Max (max) are level on quality (64.2 vs 64.2).
- GPT-5.4 (xhigh) is stronger in agents & tools, coding, instruction following, long context, maths.
- Qwen3.8 Max (max) is stronger in composite, knowledge, multimodal, reasoning, human preference.
- Qwen3.8 Max (max) is 1.9× cheaper ($3.00 vs $5.63 per 1M blended).
- GPT-5.4 (xhigh) streams 2.5× faster (91 vs 37 tokens per second).
| Metric | GPT-5.4 (xhigh) | Qwen3.8 Max (max) |
|---|---|---|
| BenchLeader Index | 64.2 | 64.2 |
| Agents & tools score | 63.6 | 58.2 |
| Coding score | 64.7 | 64.6 |
| Composite score | 66.3 | 70.7 |
| Instruction following score | 72.1 | – |
| Knowledge score | 59.9 | 62.7 |
| Long context score | 66.3 | 65.4 |
| Maths score | 65.4 | 65.1 |
| Multimodal score | 61.6 | 66.2 |
| Reasoning score | 69.5 | 72.2 |
| Human preference score | – | 68.0 |
| Blended price $/M | $5.63 | $3.00 |
| Output speed | 91 tok/s | 37 tok/s |
| Time to first answer | 154.5 s | 58.9 s |
| Context window | 1.1M | 1M |
| GPQA Diamond | 93.3% | – |
| FrontierMath Tiers 1–3 | 78.6% | – |
| FrontierMath Tier 4 | 49.0% | – |
| OTIS Mock AIME | 95.3% | – |
| SimpleQA Verified | 45.1% | – |
| Humanity's Last Exam | 36.2% | – |
| Terminal-Bench | – | 27.0% |
| SciCode | 56.6% | 53.2% |
| WeirdML | 77.7% | – |
| APEX-Agents | – | 63.3% |
| ProofBench | 56.0% | 58.0% |
| GSO-Bench | 31.4% | – |
| Epoch Capabilities Index | – | 156.4 |
| LMArena Text | – | 1483 |
| LMArena Hard Prompts | – | 1504 |
| LMArena Coding | – | 1524 |
| LMArena WebDev | – | 1672 |
| LMArena Vision | – | 1314 |
| LMArena Agent | – | 2.3 |
| LiveBench | 78.0% | 78.5% |
| LiveBench Reasoning | 88.1% | 88.2% |
| LiveBench Coding | 77.5% | 72.9% |
| LiveBench Agentic Coding | 53.8% | 64.7% |
| LiveBench Mathematics | 94.2% | 91.3% |
| LiveBench Data Analysis | 79.3% | 78.4% |
| LiveBench Language | 82.6% | 79.7% |
| LiveBench Instruction Following | 70.2% | 74.1% |
| AA Intelligence Index v4.3.2 | 39.0 | 45.4 |
| IFBench | 74.0% | – |
| AA-LCR | 82.0% | 80.3% |
| MMMU-Pro | 78.4% | 82.8% |
| AA-Omniscience | 5.8 | 12.0 |
| Terminal-Bench Hard | 57.6% | – |
| GPQA Diamond (AA) | 92.0% | 92.8% |
| Humanity's Last Exam (AA) | 43.7% | 43.1% |
| SciCode (AA) | – | 53.2% |
| τ²-Bench Telecom (AA) | 87.1% | – |
| AIME (Vals) | 96.7% | – |
| LiveCodeBench | 84.1% | 87.8% |
| MMLU-Pro | 87.5% | 88.6% |
| IOI | – | 68.9% |
| LegalBench | 86.0% | 83.6% |
| CorpFin | 65.3% | 65.8% |
| TaxEval | 74.0% | 75.5% |
| MedQA | 96.1% | – |
| Terminal-Bench 2.1 (Vals) | – | 67.4% |
| SWE-bench (Vals) | 78.2% | 85.6% |
| GPQA Diamond (Vals) | 91.7% | 93.7% |
| Vals Index | – | 48.3 |
| AIME 2026 | 99.2% | – |
| HMMT February 2026 | 97.7% | – |
| MathArena Apex | 54.2% | – |
| SWE-Bench Pro | 59.1% | – |
| MCP Atlas | 70.6% | – |
| ARC-AGI-1 | 93.7% | – |
| ARC-AGI-2 | 74.0% | – |
| CritPt | 23.4% | 20.0% |
| GDPval-AA v2.1 | 37.4% | 58.6% |
| τ³-Banking (AA) | 39.6% | 51.3% |
| ITBench SRE (AA) | – | 40.3% |
| Analyst Agent (AA) | – | 45.0% |
| APEX-Agents (AA) | 33.3% | 42.4% |
| CaseLaw v2 | 63.8% | – |
| Code Migration | 35.0% | 24.0% |
| CyberBench | – | 28.6% |
| Excel Modeling Benchmark | – | 60.1% |
| Finance Agent v2 | – | 50.6% |
| Harvey's Legal Agent Benchmark | 0.0% | 10.4% |
| Legal Research Bench | – | 47.6% |
| MedCode | 41.3% | 40.7% |
| MedScribe | 77.5% | 85.0% |
| MMMU-Pro (Vals) | 87.5% | 88.0% |
| MortgageTax | 68.3% | 64.0% |
| MysteryMechanism | – | 23.9% |
| ProgramBench | 0.0% | 0.0% |
| Public Benefits Bench | – | 67.1% |
| SAGE | 43.3% | 51.3% |
| SkillsBench | 51.7% | 42.0% |
| Tax Agent Bench | – | 66.0% |
| Terminal-Bench 4.0 (Vals) | – | 34.3% |
| Terminal-Bench Science | – | 1.4% |
| Vals Multimodal Index | – | 65.4% |
| Vibe Code Bench 1-100 | – | 12.8% |
| Vibe Code Bench v1.1 | 67.4% | 64.7% |
| SWE-Bench Pro (private) | 59.1% | – |
| SWE Atlas: Codebase QnA | 40.8% | – |
| SWE Atlas: Refactoring | 44.3% | – |
| SWE Atlas: Test Writing | 44.4% | – |
| LMArena Maths | – | 1497 |
| LMArena Creative Writing | – | 1470 |
| LMArena Instruction Following | – | 1474 |
| LMArena Multi-turn | – | 1492 |
| LMArena Longer Queries | – | 1492 |
| Chess Puzzles | 44.0% | – |
| EBR-bench | 25.4% | – |
| Mystery Game Puzzles | 37.0% | – |
| CL-bench | 27.9% | – |
| CL-bench Life | 21.7% | – |
| METR Time Horizons | 74.3% | – |
| DeepSWE v1.1 | 51.8% | – |
| LMCA | 52.0% | – |
| DTBench | 94.4% | – |
| Terminal-Bench 4.0 (AA) | – | 38.9% |
| Terminal-Bench 2.1 (AA) | 78.3% | 88.8% |
| AutomationBench | – | 56.2% |
| GDP.pdf | – | 22.8% |
| MLCR | – | 20.0% |
| EnterpriseOps-Gym | – | 47.6% |
| AA-Omniscience: accuracy | 50.9% | 31.9% |
| AA-Omniscience: non-hallucination | 8.3% | 71.2% |
| AA-Briefcase v1.1 | – | 1617 |
Data as of 2026-10-11. Best configuration of each model; every score links to its source on the model pages.
GPT-5.4 vs Qwen3.8 Max: questions
- Is GPT-5.4 better than Qwen3.8 Max?
- GPT-5.4 (xhigh) and Qwen3.8 Max (max) are level on quality (64.2 vs 64.2). The BenchLeader Index combines every independent quality benchmark; Qwen3.8 Max (max) is ahead overall as of 2026-10-11, but check the category scores for your use.
- Is GPT-5.4 better than Qwen3.8 Max for coding?
- GPT-5.4 scores higher in coding (65 vs 65 on the category index, where 50 is average).
- Is GPT-5.4 better than Qwen3.8 Max for agentic tasks?
- GPT-5.4 scores higher in agentic tasks (64 vs 58 on the category index, where 50 is average).
- Which is cheaper, GPT-5.4 or Qwen3.8 Max?
- Qwen3.8 Max is cheaper: $3.00 against $5.63 per million tokens, blended at three input tokens per output token.
- Which is faster, GPT-5.4 or Qwen3.8 Max?
- GPT-5.4 streams faster: 91 against 37 output tokens per second.
- Which has the larger context window?
- GPT-5.4 accepts more context: 1.1M against 1M tokens.