Description

An agent SDK and loop harness that enables models to solve complex reasoning tasks via token-efficient context engineering and iterative tool use. By managing environment state in external logs rather than vision-only harnesses, it allowed a model to achieve a near-perfect 99.95% score on the ARC-AGI-3 intelligence benchmark using shell and Python script execution.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.