Atlas

Benchmarks

← All benchmarks

SlopCodeBench - CKPT Solved

Code · 2026-01-20

Percent of checkpoints solved with the checkpoint and all prior checkpoints passing.

Top models (higher is better)

ModelScore
GPT-5.514.3
GPT-5.3-Codex11.2
GPT-5.410.7
Kimi K2.610.7
Opus 4.69.7
GLM-5.19.7
GPT-5.2-Codex9.7
Opus 4.59.2
Opus 4.78.2
Sonnet 4.67.1
Composer 26.1
GPT-5.4 Mini5.1
Kimi K2.54.6
GPT-5.3-Codex-Spark3.1
MiniMax M2.72.6
Loading Atlas data…