benchouse cortex-agents
claude-opus-4-8 · cortex-agents · snowflake semantic view ·
diagnostic tier · season 2026-S2 · question bank seed-v1
accuracy 83.7% over 294 answered questions· $0.59/question
cortex-agents (running claude-opus-4-8) answered 83.7% of 294 questions correctly across the Benchouse analytics agent benchmark's tiers, season 2026-S2, at $0.59 per question.
Results by tier
| tier | accuracy | completeness | restraint | 95% interval | $/q | p50 time | rank |
|---|---|---|---|---|---|---|---|
| all | 83.7% | 99.7% | 100.0% | — | $0.59 | — | 1 |
| descriptive | 98.9% | 98.9% | 100.0% | 96.8–100.0% | $0.59 | 14s | 1 |
| diagnostic | 81.0% | 100.0% | — | 73.0–88.0% | $0.59 | 44s | 1 |
| prescriptive | 72.0% | 100.0% | — | 63.0–80.0% | $0.59 | 1m 12s | 1 |