Across 33 shared benchmarks, DeepSeek-R1 scores higher on 3 and Qwen3.5 122B A10B on 30. The widest gap is τ²-Bench Telecom (AA run), where Qwen3.5 122B A10B scores 93.6 against 11.4. Tracked API pricing per million tokens: DeepSeek-R1 $1.35 in / $3.00 out, Qwen3.5 122B A10B $0.40 in / $3.20 out.
| Benchmark | DeepSeek-R1 | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 3.1 | 21.3 |
| AA Intelligence | 18.6 | 32.8 |
| AA-LCR | 56 | 70.3 |
| AA-Omniscience | -31.3 | -41.5 |
| Artificial Analysis Coding Index | 24.6 | 45.7 |
| browsecomp | 8.9 | 63.8 |
| browsecomp_zh | 35.7 | 69.9 |
| C-Eval | 91.8 | 91.9 |
| critpt | 0.6 | 0.9 |
| gdpval | 1.5 | 24.3 |
| GPQA Diamond | 81 | 86.6 |
| HLE | 17.7 | 47.5 |
| HMMT 2025 | 79.4 | 90.3 |
| IFBench | 39 | 76.1 |
| ifeval | 83.3 | 93.4 |
| longbench_v2 | 58.3 | 60.2 |
| mmlu_prox | 75.5 | 82.2 |
| mmlu_redux | 93.4 | 94 |
| MMLU-Pro | 85 | 86.7 |
| multichallenge | 45 | 61.5 |
| OmniScience Accuracy | 30.7 | 24.4 |
| OmniScience Non-Hallucination | 10.5 | 12.9 |
| scicode | 35.7 | 42 |
| SWE-bench Verified | 57.6 | 72 |
| TauBench V3 - Banking | 6.4 | 15.3 |
| Terminal-Bench 2.1 | 19.1 | 47.6 |
| Terminal-Bench Hard | 6.1 | 31.1 |
| vectara_answer_rate | 97 | 99.8 |
| vectara_avg_summary_length | 93.5 | 86.4 |
| vectara_factual_consistency | 88.7 | 88.8 |
| vectara_hallucination_rate ↓ | 11.3 | 11.2 |
| τ²-Bench Telecom (AA run) | 11.4 | 93.6 |
| τ³-Bench | 6.4 | 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 (direct, direct), otherwise the lowest tracked offer.