Atlas

Benchmarks

← All benchmarks

Deep Research Bench — Find Number

Search · 2025-05-06

The DRB Find Number category asks an agent to locate a reliable, known number on the frozen web corpus.

Top models (higher is better)

ModelScore
Opus 4.60.7
Sonnet 4.60.7
Opus 4.50.7
Gemini 3 Flash Preview0.7
Haiku 4.50.7
GPT-5.50.7
Opus 4.80.7
Grok 40.7
Sonnet 4.50.7
Gemini 3.1 Pro Preview0.7
GPT-50.6
Opus 4.10.6
Sonnet 40.6
Gemini 3 Pro Preview0.6
Gemini 2.5 Pro0.6
Opus 40.6
o30.6
GPT-5.10.6
GPT-5.20.6
GPT-5.40.6
Gemini 3.1 Flash-Lite Preview0.6
GPT-5.4 Mini0.6
Loading Atlas data…