Atlas

Benchmarks

← All benchmarks

CL-Bench Exploitable Poker Reward

Agents · 2026-05-04

Mean cumulative reward across 120 sequential instances of the CL-Bench exploitable poker task.

Top models (higher is better)

ModelScore
Sonnet 4.6343
GPT-5.4144
Opus 4.7117
Gemini 3 Flash Preview94.8
Gemini 3.1 Pro Preview76.4
Loading Atlas data…