GLM-5.2 is the best cost-per-intelligence model on the market right now.

~80% cheaper than Claude, with ~95% of the intelligence of Opus 4.8.
Insane value.
Most people don't realize how huge a leap this really is.
GLM-5.2: $5.80 per million tokens, ~95% of frontier intelligence
GPT-5.5: ~$9.70, 92% intelligence
Opus 4.8: ~$23, 98% intelligence
Fable 5: ~$30, 100% intelligence
Plot it on a cost-vs-intelligence chart, and GLM-5.2 sits exactly where you want to be: high intelligence, low cost.
This is the first open-source model to hit frontier-level performance at a fraction of the price.
However, I'm still using Claude and other frontier models regularly with a "barbell" approach (which most people should be doing).
My barbell strategy:
Opus 4.8 for the initial 10% planning.
GLM-5.2 for the cheap 80% (grunt work, first drafts, repetitive tasks).
Opus 4.8 for the last-step verification.