The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. No benchmark-specific protocol is stated.
ProgramBench (Almost Solved) · GLM-5.3模型卡;Almost Solved · Reported score · The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. No benchmark-specific protocol is stated.
Regular display scope. Rankings are grouped by metric and evaluation protocol. Recorded tool access, budgets, and frameworks are shown separately. Scores remain available for reference when conditions are unspecified.
Reported score
The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. No benchmark-specific protocol is stated.
ProgramBench (Almost Solved) · GLM-5.3模型卡;Almost Solved · Reported score · The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. No benchmark-specific protocol is stated.