Skip to content
mager-bench1.3

Saved attempts

Every v1.3 attempt, including failed calls. A missing answer never becomes a zero.

No v1.3 attempts yet

The new tasks are validated with author-written reference programs, not model responses. GPT-6 Sol’s last subscription attempt was rejected on v1.2 and remains in the archive.

View v1.2 and its saved attempt

View earlier benchmark versions