Qwen3.8 Max vs Step 5 Preview
Qwen3.8 Max 
Step 5 Preview Verdict
- Qwen3.8 Max (max) leads on quality: 64.2 vs 62.4.
- Qwen3.8 Max (max) is stronger in agents & tools, human preference, maths, multimodal.
- Step 5 Preview is stronger in coding, composite, knowledge, long context.
- Step 5 Preview is 2.1× cheaper ($1.43 vs $3.00 per 1M blended).
- Step 5 Preview streams 2.4× faster (87 vs 37 tokens per second).
| Metric | Qwen3.8 Max (max) | Step 5 Preview |
|---|---|---|
| BenchLeader Index | 64.2 | 62.4 |
| Agents & tools score | 58.2 | 47.2 |
| Coding score | 64.6 | 65.6 |
| Composite score | 70.7 | 79.1 |
| Human preference score | 68.0 | 63.6 |
| Knowledge score | 62.7 | 69.1 |
| Long context score | 65.4 | 69.5 |
| Maths score | 65.1 | 61.1 |
| Multimodal score | 66.2 | 60.0 |
| Reasoning score | 72.2 | 72.2 |
| Blended price $/M | $3.00 | $1.43 |
| Output speed | 37 tok/s | 87 tok/s |
| Time to first answer | 58.9 s | 25.9 s |
| Context window | 1M | 1M |
| Terminal-Bench | 27.0% | – |
| SciCode | 53.2% | 58.9% |
| APEX-Agents | 63.3% | – |
| ProofBench | 58.0% | 42.0% |
| Epoch Capabilities Index | 156.4 | – |
| LMArena Text | 1483 | 1447 |
| LMArena Hard Prompts | 1504 | 1479 |
| LMArena Coding | 1524 | 1503 |
| LMArena WebDev | 1672 | – |
| LMArena Vision | 1314 | 1267 |
| LMArena Agent | 2.3 | 0 |
| 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% | – |
| LiveBench Instruction Following | 74.1% | – |
| AA Intelligence Index v4.3.2 | 45.4 | 43.7 |
| AA-LCR | 80.3% | 88.3% |
| MMMU-Pro | 82.8% | 76.4% |
| AA-Omniscience | 12.0 | 16.4 |
| GPQA Diamond (AA) | 92.8% | – |
| Humanity's Last Exam (AA) | 43.1% | 46.5% |
| SciCode (AA) | 53.2% | 58.9% |
| LiveCodeBench | 87.8% | – |
| MMLU-Pro | 88.6% | – |
| IOI | 68.9% | – |
| 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 | 48.3 | – |
| CritPt | 20.0% | 20.9% |
| GDPval-AA v2.1 | 58.6% | 54.2% |
| τ³-Banking (AA) | 51.3% | – |
| ITBench SRE (AA) | 40.3% | 55.6% |
| Analyst Agent (AA) | 45.0% | 35.0% |
| APEX-Agents (AA) | 42.4% | 38.0% |
| Code Migration | 24.0% | – |
| CyberBench | 28.6% | – |
| Excel Modeling Benchmark | 60.1% | – |
| Finance Agent v2 | 50.6% | – |
| Harvey's Legal Agent Benchmark | 10.4% | – |
| Legal Research Bench | 47.6% | – |
| MedCode | 40.7% | – |
| MedScribe | 85.0% | – |
| MMMU-Pro (Vals) | 88.0% | – |
| MortgageTax | 64.0% | – |
| MysteryMechanism | 23.9% | – |
| ProgramBench | 0.0% | – |
| Public Benefits Bench | 67.1% | – |
| SAGE | 51.3% | – |
| SkillsBench | 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 | 64.7% | – |
| LMArena Maths | 1497 | 1478 |
| LMArena Creative Writing | 1470 | 1406 |
| LMArena Instruction Following | 1474 | 1449 |
| LMArena Multi-turn | 1492 | 1453 |
| LMArena Longer Queries | 1492 | 1462 |
| Terminal-Bench 4.0 (AA) | 38.9% | 33.3% |
| Terminal-Bench 2.1 (AA) | 88.8% | – |
| AutomationBench | 56.2% | 51.0% |
| GDP.pdf | 22.8% | 14.8% |
| MLCR | 20.0% | 16.7% |
| EnterpriseOps-Gym | 47.6% | 47.2% |
| AA-Omniscience: accuracy | 31.9% | 41.5% |
| AA-Omniscience: non-hallucination | 71.2% | 57.0% |
| AA-Briefcase v1.1 | 1617 | 1425 |
Data as of 2026-10-11. Best configuration of each model; every score links to its source on the model pages.
Qwen3.8 Max vs Step 5 Preview: questions
- Is Qwen3.8 Max better than Step 5 Preview?
- Qwen3.8 Max (max) leads on quality: 64.2 vs 62.4. 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 Qwen3.8 Max better than Step 5 Preview for coding?
- Step 5 Preview scores higher in coding (66 vs 65 on the category index, where 50 is average).
- Is Qwen3.8 Max better than Step 5 Preview for agentic tasks?
- Qwen3.8 Max scores higher in agentic tasks (58 vs 47 on the category index, where 50 is average).
- Which is cheaper, Qwen3.8 Max or Step 5 Preview?
- Step 5 Preview is cheaper: $1.43 against $3.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Qwen3.8 Max or Step 5 Preview?
- Step 5 Preview streams faster: 87 against 37 output tokens per second.
- Which has the larger context window?
- Both accept 1M tokens of context.