Benchmarks › AA Omniscience
AA Omniscience
Answer knowledge questions or abstain instead of guessing
67.2top score · Claude Fable 5.1
318models ranked
215open-weights models
Accuracyhigher is better
Oct 20, 2026last updated
How the test works
1
Task
6,000 questions
- What the model doesAnswer knowledge questions or abstain instead of guessing
- Kinds of tasksKnowledge questions
- In → outText → Answer, abstention
2
Run
1 run per task
- Token budgetNon-reasoning: up to 16,384 output tokens, adjusted to model limits; reasoning: model-specific maximum
- Temperature0 for non-reasoning; 0.6 for reasoning, unless the model lab recommends otherwise
3
Score
Accuracy
- Graded byCorrect / incorrect / partial / not attempted, judged by GPT-5.6 Luna (medium)
- Better isHigher
- Test setPrivate questions; a public sample is separately released
Sources: artificialanalysis.ai
AA Omniscience · Accuracy leaderboardtop 10 of 318
Full leaderboard with filters →| Model | AA Omniscience · Accuracy | Per task | Price | Speed | Supports | Context |
|---|---|---|---|---|---|---|
67.2max | $7.63 | $10 / $50 | 69 | 1M | ||
66.2max | $5.98 | $4 / $20 | 96 | 1M | ||
65.4max | $8.75 | $10 / $50 | 67 | 1M | ||
62.6max | $3.26 | $10 / $50 | 69 | 1.05M | ||
62.1max | $0.72 | $2 / $10 | 52 | 1.05M | ||
60.9max | $5.86 | $5 / $25 | 55 | 1M | ||
59.4max | $1.99 | $4 / $20 | 114 | 1.05M | ||
58xhigh | $2.63 | $5 / $30 | 113 | 1.05M | ||
56.2thinking | $0.22 | $0.15 / $0.60 | 105 | 1M | ||
54.9thinking | $1.30 | $2 / $12 | 115 | 1M |
Versions and boards on Model Menu
| Board | Models |
|---|---|
AA Omniscience · AccuracyShown above | 321 |
| 321 | |
| 321 |
AA Omniscience questions
What is AA Omniscience?›
Answer knowledge questions or abstain instead of guessing. It has 6,000 questions. Created by Artificial Analysis.
Which model is best on AA Omniscience?›
Claude Fable 5.1 leads AA Omniscience · Accuracy with 67.2, ahead of Claude Opus 5.5 (66.2) and Claude Fable 5 (65.4).
What is the best open-weights model on AA Omniscience?›
DeepSeek V4 Pro 0813 with 49.1, then Kimi K3 (47.6).
How is AA Omniscience scored?›
Accuracy; correct / incorrect / partial / not attempted, judged by gpt-5.6 luna (medium); 1 run per task per task.
Which versions of AA Omniscience are there?›
AA Omniscience · Accuracy (321 models), AA Omniscience · Hallucination Rate (321 models), AA Omniscience Index (321 models).