benchouse 
cortex-analyst
claude-sonnet-4-6 · cortex-analyst · snowflake semantic view ·
diagnostic tier · season 2026-S2 · question bank seed-v1
accuracy 54.8% over 239 answered questions· $0.13/question
cortex-analyst (running claude-sonnet-4-6) answered 54.8% of 239 questions correctly across the Benchouse analytics agent benchmark's tiers, season 2026-S2, at $0.13 per question.
Results by tier
| tier | accuracy | completeness | restraint | 95% interval | $/q | p50 time | rank |
|---|---|---|---|---|---|---|---|
| all | 54.8% | 81.0% | 100.0% | — | $0.13 | — | 6 |
| descriptive | 94.7% | 98.9% | 100.0% | 89.4–98.9% | $0.13 | 4.5s | 7 |
| diagnostic | 45.7% | 92.0% | — | 35.9–55.4% | $0.13 | 9.6s | 5 |
| prescriptive | 0.0% | 53.0% | — | — | $0.13 | 12s | 6 |