| Benchmark | Claude Opus 4.7 | DeepSeek-V3.2 |
|---|---|---|
| AA Agentic Index | 64.6 | 39.8 |
| AA Intelligence | 55 | 32.8 |
| AA-LCR | 75.3 | 70.7 |
| AA-Omniscience | 27.3 | -22.5 |
| aime_2026 | 95.8 | 95.1 |
| arc_agi_1 | 92 | 57 |
| ARC-AGI-2 | 75.8 | 4 |
| arena_text_factuality | 1474 | 1433 |
| Artificial Analysis Coding Index | 73.6 | 44.2 |
| browsecomp | 79.8 | 51.4 |
| coding_arena_elo | 1557 | 1368 |
| critpt | 12 | 2.9 |
| cybergym | 73.8 | 17.3 |
| frontiermath_tier_4 | 22.9 | 2.1 |
| gaia_overall | 28 | 76.7 |
| gdpval | 58.6 | 18.8 |
| GPQA Diamond | 94.2 | 84 |
| HLE | 54.7 | 40.8 |
| HLE (with tools) | 54.7 | 40.8 |
| hmmt_feb_2026 | 93.9 | 84.1 |
| IFBench | 58.6 | 61 |
| MCP Atlas | 79.1 | 62.2 |
| OmniScience Accuracy | 48.9 | 33 |
| OmniScience Non-Hallucination | 57.7 | 17.3 |
| scicode | 54.5 | 39 |
| simpleqa_verified | 50.6 | 27.5 |
| SWE-bench Multilingual | 80.5 | 70.2 |
| SWE-bench Pro | 64.3 | 15.6 |
| SWE-bench Verified | 87.6 | 73.1 |
| TauBench V3 - Banking | 34.6 | 18.8 |
| Terminal-Bench 2.0 | 69.4 | 46.4 |
| Terminal-Bench 2.1 | 83.1 | 46.8 |
| Terminal-Bench Hard | 54.5 | 35.6 |
| toolathlon | 59.3 | 35.2 |
| vectara_answer_rate | 98 | 92.6 |
| vectara_avg_summary_length | 149.1 | 62 |
| vectara_factual_consistency | 88 | 93.7 |
| vectara_hallucination_rate ↓ | 12 | 6.3 |
| τ²-Bench Telecom (AA run) | 88.6 | 90.6 |
| τ³-Bench | 28.9 | 69.2 |
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.