AA-Omniscience: accuracy
The share of AA-Omniscience questions answered correctly, before any penalty for confident wrong answers.
As of 22 Sept 2026, Claude Fable 5.1 leads AA-Omniscience: accuracy on BenchLeader with 67.2%, ahead of Claude Fable 5.1 at 66.2%, across 501 model configurations with a published result.
- Published by
- Artificial Analysis
- Category
- Knowledge
- Index weight
- Reference only
- Models
- 501
- Data as of
- 22 Sept 2026
Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.
- 1Claude Fable 5.1 (thinking)67.2%
- 2Claude Fable 5.1 (xhigh)66.2%
- 3Claude Opus 5.5 (thinking)66.2%
- 4Claude Opus 5.5 (xhigh)65.4%
- 5Claude Fable 5 (thinking)65.3%
- 6Claude Fable 5.1 (high)64.9%
- 7Claude Opus 5.5 (high)64.6%
- 8Claude Opus 5.5 (medium)64.5%
- 9Claude Opus 5.5 (low)63.5%
- 10Claude Fable 5.1 (medium)63.1%
- 11GPT-6 Astra (max)62.6%
- 12GPT-6 Astra (xhigh)61.9%
- 13GPT-6 Astra (high)61.1%
- 14Claude Opus 5 (max)60.9%
- 15GPT-6 Astra (medium)60.6%
501 of 501
| # | |||
|---|---|---|---|
| 1 |  Claude Fable 5.1thinkingAnthropic | 67.2% | 71.0 |
| 2 | Claude Fable 5.1xhighAnthropic | 66.2% | 71.1 |
| 3 | NewClaude Opus 5.5thinkingAnthropic | 66.2% | 70.7 |
| 4 | NewClaude Opus 5.5xhighAnthropic | 65.4% | 70.3 |
| 5 | Claude Fable 5thinkingAnthropic | 65.3% | 70.3 |
| 6 | Claude Fable 5.1highAnthropic | 64.9% | 71.0 |
| 7 | NewClaude Opus 5.5highAnthropic | 64.6% | 69.9 |
| 8 | NewClaude Opus 5.5mediumAnthropic | 64.5% | 69.3 |
| 9 | NewClaude Opus 5.5lowAnthropic | 63.5% | 65.2 |
| 10 | Claude Fable 5.1mediumAnthropic | 63.1% | 69.1 |
| 11 |  GPT-6 AstramaxOpenAI | 62.6% | 71.7 |
| 12 | GPT-6 AstraxhighOpenAI | 61.9% | 70.5 |
| 13 | GPT-6 AstrahighOpenAI | 61.1% | 71.3 |
| 14 | Claude Opus 5maxAnthropic | 60.9% | 69.5 |
| 15 | GPT-6 AstramediumOpenAI | 60.6% | 69.2 |
| 16 | Claude Fable 5.1lowAnthropic | 60.2% | 66.7 |
| 17 | GPT-6 AstralowOpenAI | 59.5% | 67.6 |
| 18 | Claude Opus 5xhighAnthropic | 59.5% | 69.3 |
| 19 | GPT-5.6 SolmaxOpenAI | 59.4% | 68.5 |
| 20 | Claude Opus 5highAnthropic | 58.9% | 69.8 |
| 21 | GPT-5.6 SolxhighOpenAI | 58.8% | 67.9 |
| 22 | GPT-5.6 SolhighOpenAI | 58.4% | 67.4 |
| 23 | GPT-5.5xhighOpenAI | 58.0% | 67.2 |
| 24 | GPT-5.6 SolmediumOpenAI | 57.8% | 65.5 |
| 25 | GPT-5.6 SollowOpenAI | 57.2% | 63.2 |
| 26 | Claude Opus 5mediumAnthropic | 57.1% | 66.4 |
| 27 | GPT-5.5highOpenAI | 57.0% | 66.6 |
| 28 | GPT-5.5mediumOpenAI | 56.8% | 64.0 |
| 29 | Claude Opus 5lowAnthropic | 56.0% | 62.8 |
| 30 |  Gemini 3 ProhighGoogle | 55.8% | 60.7 |
| 31 | Gemini 3.7 FlashhighGoogle | 55.3% | 64.3 |
| 32 | Gemini 3.1 ProGoogle | 54.9% | 63.2 |
| 33 | GPT-5.5lowOpenAI | 54.9% | 60.8 |
| 34 | Gemini 3.8 FlashhighGoogle | 54.6% | 64.3 |
| 35 | Gemini 3.7 FlashmediumGoogle | 54.0% | 64.4 |
| 36 | Gemini 3.7 FlashlowGoogle | 53.6% | 61.6 |
| 37 | Gemini 3 FlashthinkingGoogle | 53.4% | 60.0 |
| 38 | Gemini 3.8 FlashmediumGoogle | 53.0% | 64.6 |
| 39 | GPT-5.3-CodexxhighOpenAI | 52.9% | 64.9 |
| 40 | Gemini 3.8 FlashlowGoogle | 52.2% | 60.2 |
| 41 |  Muse Spark 1.1xhighMeta | 52.0% | 61.4 |
| 42 |  Grok 4.5highSpaceXAI | 51.5% | 60.6 |
| 43 | Grok Build 0.1SpaceXAI | 51.5% | 54.5 |
| 44 | Gemini 3.5 FlashhighGoogle | 51.4% | 63.4 |
| 45 | Gemini 3.5 FlashmediumGoogle | 51.0% | 63.9 |
| 46 | GPT-5.4xhighOpenAI | 50.9% | 65.1 |
| 47 | Gemini 3.6 FlashhighGoogle | 50.0% | 61.5 |
| 48 | Muse SparkMeta | 49.6% | 65.4 |
| 49 |  DeepSeek V4 PromaxDeepSeekopen ↗ | 49.1% | 63.7 |
| 50 | Claude Opus 4.7maxAnthropic | 48.9% | 63.7 |
| 51 | Claude Opus 4.8maxAnthropic | 48.8% | 63.5 |
| 52 | GPT-5.6 Solno reasoningOpenAI | 48.7% | 54.8 |
| 53 | Gemini 3 ProlowGoogle | 48.3% | 54.4 |
| 54 | Grok 4.6highSpaceXAI | 48.2% | 64.2 |
| 55 | GPT-5.4lowOpenAI | 47.9% | 60.5 |
| 56 | NewGrok 4.7highSpaceXAI | 47.8% | – |
| 57 |  Kimi K3maxMoonshot AIopen ↗ | 47.6% | 66.7 |
| 58 | NewGrok 4.7xhighSpaceXAI | 47.5% | 61.9 |
| 59 | Claude Opus 4.6maxAnthropic | 47.0% | 58.5 |
| 60 | GPT-5.6 TerramaxOpenAI | 46.8% | 64.8 |
Cite as: BenchLeader, “AA-Omniscience: accuracy leaderboard”, https://www.benchleader.com/benchmarks/aa_omniscience_accuracy, data as of 22 Sept 2026.
AA-Omniscience: accuracy: questions
- What does AA-Omniscience: accuracy measure?
- The share of AA-Omniscience questions answered correctly, before any penalty for confident wrong answers. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads AA-Omniscience: accuracy?
- Claude Fable 5.1 leads AA-Omniscience: accuracy with 67.2% as of 22 Sept 2026, ahead of Claude Fable 5.1 at 66.2%.
- How many models have AA-Omniscience: accuracy results?
- 501 model configurations have a AA-Omniscience: accuracy result on BenchLeader, all taken from Artificial Analysis.
- Who runs AA-Omniscience: accuracy and how often is it updated?
- AA-Omniscience: accuracy is published by Artificial Analysis. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does AA-Omniscience: accuracy count toward the BenchLeader Index?
- No. AA-Omniscience: accuracy is shown for reference but left out of the composite index.