It's not perfect, but there is something http://Z.ai is doing in post-training that makes its last model qualitatively better (GLM 5.1) than competitors.
With other models I feel like I'm the environment being maxxed: everything goes -- lying / cheating / deceiving.
Been using GLM 5.1 as daily driver and wanted to switch to another tier-equivalent model for code review: it's like wrestling match. You do get stuff out, but every step is like... George Bernard Shaw's quote:
"Never wrestle with pigs. You both get dirty and the pig likes it."
It's not perfect, but there is something http://Z.ai is doing in post-training that makes its last model qualitatively better (GLM 5.1) than competitors.
With other models I feel like I'm the environment being maxxed: everything goes -- lying / cheating / deceiving.
It's not perfect, but there is something http://Z.ai is doing in post-training that makes its last model qualitatively better (GLM 5.1) than competitors.
With other models I feel like I'm the environment being maxxed: everything goes -- lying / cheating / deceiving.
Users praise GLM 5.1 as their daily driver because it outperforms rival models in tasks like code review thanks to superior post-training.