Artificial Analysis
AA-AnalystAgent
Agents work through folders of spreadsheets and documents to answer quantitative questions that business and data analysts face. The emphasis is on judgement and method. It spans 14 business and scientific domains and is not finance-specific.
About this benchmark
- Size
- 80 questions
- Public split
- none (only the example tasks shown on the page)
- Grading
- Each question runs 5 times. Headline pass^5 is the share answered correctly in all 5 runs; pass@1 and pass@5 are also reported. How individual answers are judged is not stated on the page.
- Licence
- Not stated
- Submission
- Not stated
- Products allowed
- Not stated
- What it does not cover
- Not focused on finance: no filings research, markets or advice.
Results
3 modelsEach score shows who reported it: the benchmark's authors (quoted from their paper), the benchmark's official board, or Moneda (run with the benchmark's own grader and rules). We do not copy other aggregators.
AA-AnalystAgent
pass^5 %, higher is better. 3 models.
Updated date not stated
AA-AnalystAgent
pass^5 %, higher is better. 3 models.
Updated date not stated
- 1Gemini 3.7 Flash (High)60
- 2Claude Fable 5.1 (Max, Default Fallback)57.5
- 2Claude Sonnet 5.5 (Max, Default Fallback)57.5
| 1 | Gemini 3.7 Flash (High) | 60 | Stirrup harness | Artificial Analysis evaluation page Official board, date not stated | |
|---|---|---|---|---|---|
| 2 | Claude Fable 5.1 (Max, Default Fallback) | Anthropic | 57.5 | Stirrup harness | Artificial Analysis evaluation page Official board, date not stated |
| 2 | Claude Sonnet 5.5 (Max, Default Fallback) | Anthropic | 57.5 | Stirrup harness | Artificial Analysis evaluation page Official board, date not stated |
Papers and links
Last checked 1 Nov 2026. Something out of date or wrong? Send a correction.