GLM 5.2 reached nearly $8,000 while GPT-5.5 trailed under $4,000.
This is a remarkable graph that reinforces the same point as autoresearch results, WeirdML, "takeout delivery bench", and all other long multi-turn scenarios. Notice: *GLM-5.2 isn't any better than 5-5.1 for the first ≈150 days*. But it learns in-context. GPT-5 takes 300 days.
GLM 5.2 is 2nd in Vending-Bench.
Each GLM release has improved at a remarkably steady pace: a linear fit of R²=0.99, with almost $1k better per month.
Users praise GLM-5.2's Vending-Bench gains as the result of patient long-term accumulation and in-context learning rather than rushed development.
No Digg Deeper questions have been answered for this story yet.
GLM 5.2 is 2nd in Vending-Bench.
Each GLM release has improved at a remarkably steady pace: a linear fit of R²=0.99, with almost $1k better per month.
@teortaxesTex 这就叫厚积薄发时间才是硬核门槛
@andonlabs Oh, great! Maybe this model would deserve the vacant spot in the DJ booth of the Grok'n Roll station?
