Across 27 shared benchmarks, DeepSeek-V4-Pro scores higher on 24 and Qwen3.5 122B A10B on 3. The widest gap is AA Agentic Index, where DeepSeek-V4-Pro scores 63.3 against 21.3. Tracked API pricing per million tokens: DeepSeek-V4-Pro $0.66 in / $1.98 out, Qwen3.5 122B A10B $0.40 in / $3.20 out.
| Benchmark | DeepSeek-V4-Pro | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 63.3 | 21.3 |
| AA Intelligence | 53 | 32.8 |
| AA-LCR | 70 | 70.3 |
| AA-Omniscience | -10.6 | -41.5 |
| Artificial Analysis Coding Index | 59.4 | 45.7 |
| browsecomp | 83.4 | 63.8 |
| coding_arena_elo | 1582 | 1358 |
| critpt | 13 | 0.9 |
| gdpval | 49 | 24.3 |
| GPQA Diamond | 90.5 | 86.6 |
| HLE | 48.2 | 47.5 |
| IFBench | 76.5 | 76.1 |
| MMLU-Pro | 87.5 | 86.7 |
| OmniScience Accuracy | 43 | 24.4 |
| OmniScience Non-Hallucination | 12.2 | 12.9 |
| scicode | 50 | 42 |
| SWE-bench Verified | 80.6 | 72 |
| TauBench V3 - Banking | 30.1 | 15.3 |
| Terminal-Bench 2.0 | 67.9 | 49.4 |
| Terminal-Bench 2.1 | 72.1 | 47.6 |
| Terminal-Bench Hard | 46.2 | 31.1 |
| vectara_answer_rate | 97.2 | 99.8 |
| vectara_avg_summary_length | 153.8 | 86.4 |
| vectara_factual_consistency | 91.4 | 88.8 |
| vectara_hallucination_rate ↓ | 8.6 | 11.2 |
| τ²-Bench Telecom (AA run) | 96.2 | 93.6 |
| τ³-Bench | 25.8 | 13.6 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (deepseek-official, alibaba-official), otherwise the lowest tracked offer.