Across 24 shared benchmarks, DeepSeek-V4-Flash scores higher on 21 and Qwen3.5 122B A10B on 3. The widest gap is AA Agentic Index, where DeepSeek-V4-Flash scores 61.3 against 21.3. DeepSeek-V4-Flash is the cheaper of the two on tracked API pricing ($0.22 against $0.40 per million input tokens).
| Benchmark | DeepSeek-V4-Flash | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 61.3 | 21.3 |
| AA Intelligence | 51.8 | 32.8 |
| AA-LCR | 74.3 | 70.3 |
| AA-Omniscience | -14.3 | -41.5 |
| Artificial Analysis Coding Index | 69.1 | 45.7 |
| browsecomp | 73.2 | 63.8 |
| coding_arena_elo | 1431 | 1358 |
| critpt | 16.6 | 0.9 |
| gdpval | 52.9 | 24.3 |
| GPQA Diamond | 90.8 | 86.6 |
| HLE | 38.6 | 47.5 |
| IFBench | 79.2 | 76.1 |
| MMLU-Pro | 86.4 | 86.7 |
| MMMU-Pro | 74.8 | 76.9 |
| OmniScience Accuracy | 40.4 | 24.4 |
| OmniScience Non-Hallucination | 15.6 | 12.9 |
| scicode | 49.9 | 42 |
| SWE-bench Verified | 79 | 72 |
| TauBench V3 - Banking | 39.4 | 15.3 |
| Terminal-Bench 2.0 | 56.9 | 49.4 |
| Terminal-Bench 2.1 | 82.7 | 47.6 |
| Terminal-Bench Hard | 38.6 | 31.1 |
| τ²-Bench Telecom (AA run) | 95.6 | 93.6 |
| τ³-Bench | 22.9 | 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.