← Claude Sonnet 5 · API · High effort (default)
Evidenceacceptedindependent_lab
AA-Omniscience
Artificial Analysis — AA-Omniscience (Claude Sonnet 5 max)
ev_2026_07_sonnet5_aa_omniscience
Result
15.32
Confidence A
- Tested
- 2026-07-01
- Runs
- 1
Configuration
How it was measured
- Exact model
- claude-sonnet-5
- Reasoning
- max
- Agent harness
- none
- Tool access
- AA default knowledge probe (not web-search agent)
- Prompting
- Artificial Analysis Omniscience protocol
PRIMARY independent WIS signal — calibration / uncertainty management via AA-Omniscience (accuracy vs hallucination tradeoff).
Caveats
Known limitations
- Omniscience Index is not percent accuracy; display accuracy + hallucination rate.
- Max effort config.
- Positive index is relatively strong in a field where many models score ≤0.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.95
quality1.00
recency1.00
health0.85
comparability0.90
reliability0.85
Benchmark
AA-Omniscience
Knowledge reliability benchmark that jointly scores accuracy and hallucination / overconfidence. Rewards calibrated abstention rather than pure guess accuracy.
Benchmark source →