Running the same triangle test on the new GLM-5.3.
Comparing it with Claude Opus 5 and GPT-5.6 Sol.
Opus 5 charges $15 in and $75 out per million.
Grok 4.5 charges $2 and $6. That is 12.5X on every word your agent writes, and 91% of what mine wrote last month was extraction and formatting that never needed a $75 lane here is the whole bill, task to invoice: 1,140 agent calls
Claude Opus 5 is nowadays sooo fucking dumb, holy fuck.
Claude, fuck off. I know you want to increase profits ahead of the IPO, but the current models suck. On benchmarks, Grok 4.6 is a worse model, but somehow it actually gets shit done
I tested Opus 5 Ultra Code vs GPT-5.6 Extra High on a real Swift bug — the results surprised me 👀.
You have reached the end of the archive
All of Claude Opus 5