Claude Sonnet 5.5 vs Deepseek v4 Pro High
Claude Sonnet 5.5 
Deepseek v4 Pro High Verdict
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 60.0.
- Claude Sonnet 5.5 (max) is stronger in agents & tools, coding, composite, knowledge, long context, maths, reasoning.
- Deepseek v4 Pro High (high) is stronger in human preference.
| Metric | Claude Sonnet 5.5 (max) | Deepseek v4 Pro High (high) |
|---|---|---|
| BenchLeader Index | 67.6 | 60.0 |
| Agents & tools score | 69.5 | – |
| Coding score | 69.8 | 57.2 |
| Composite score | 72.6 | – |
| Knowledge score | 65.4 | – |
| Long context score | 66.6 | – |
| Maths score | 72.9 | 65.2 |
| Reasoning score | 81.1 | 65.4 |
| Human preference score | – | 65.7 |
| Blended price $/M | $4.00 | – |
| Output speed | 141 tok/s | – |
| Time to first answer | 487.1 s | – |
| Context window | 1M | – |
| GPQA Diamond | 95.6% | 90.9% |
| FrontierMath Tiers 1–3 | 88.8% | – |
| FrontierMath Tier 4 | 80.5% | – |
| OTIS Mock AIME | 100.0% | 95.6% |
| SimpleQA Verified | 46.5% | – |
| Terminal-Bench | 61.8% | – |
| SciCode | 61.0% | 46.4% |
| WeirdML | – | 46.5% |
| APEX-Agents | 75.5% | – |
| ProofBench | 100.0% | – |
| LMArena Text | – | 1465 |
| LMArena Hard Prompts | – | 1483 |
| LMArena Coding | – | 1506 |
| LMArena WebDev | – | 1582 |
| LMArena Agent | 12 | – |
| LiveBench | 75.7% | – |
| LiveBench Reasoning | 91.6% | – |
| LiveBench Coding | 91.4% | – |
| LiveBench Agentic Coding | 56.3% | – |
| LiveBench Mathematics | 96.1% | – |
| LiveBench Data Analysis | 59.5% | – |
| LiveBench Language | 78.0% | – |
| LiveBench Instruction Following | 56.8% | – |
| AA Intelligence Index v4.3.2 | 56 | – |
| AA-LCR | 82.7% | – |
| AA-Omniscience | 32.3 | – |
| Humanity's Last Exam (AA) | 55.0% | – |
| SciCode (AA) | 61.0% | – |
| CritPt | 31.4% | – |
| GDPval-AA v2.1 | 67.0% | – |
| Analyst Agent (AA) | 57.5% | – |
| LMArena Maths | – | 1466 |
| LMArena Creative Writing | – | 1446 |
| LMArena Instruction Following | – | 1461 |
| LMArena Multi-turn | – | 1472 |
| LMArena Longer Queries | – | 1475 |
| Chess Puzzles | – | 13.0% |
| Mystery Game Puzzles | 65.0% | – |
| CL-bench Life | – | 13.5% |
| Surface Evolver Bench | – | 40.0% |
| CursorBench | 55.5% | – |
| ALE-Bench | – | 1006.1 |
| FrontierSWE | 61.9% | – |
| Terminal-Bench 4.0 (AA) | 63.6% | – |
| AutomationBench | 71.8% | – |
| GDP.pdf | 25.8% | – |
| MLCR | 75.0% | – |
| Harvey LAB | 2.8% | – |
| AA-Omniscience: accuracy | 54.0% | – |
| AA-Omniscience: non-hallucination | 53.0% | – |
| AA-Briefcase v1.1 | 1823 | – |
Data as of 2026-10-11. Best configuration of each model; every score links to its source on the model pages.
Claude Sonnet 5.5 vs Deepseek v4 Pro High: questions
- Is Claude Sonnet 5.5 better than Deepseek v4 Pro High?
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 60.0. The BenchLeader Index combines every independent quality benchmark; Claude Sonnet 5.5 (max) is ahead overall as of 2026-10-11, but check the category scores for your use.
- Is Claude Sonnet 5.5 better than Deepseek v4 Pro High for coding?
- Claude Sonnet 5.5 scores higher in coding (70 vs 57 on the category index, where 50 is average).