GLM 5.2 vs Claude Fable 5: agentic coding at a fraction of the cost
I gave GLM 5.2 and Claude Fable 5 the same real job: redesign a project plan and start implementing it. Fable 5 finished in about 9 minutes and cost me a little over $10. GLM 5.2 took about 17 minutes and cost $2.76. The open model was slower, no question. But the output was just as good for this task, and it cost roughly a quarter as much. That math has me excited about running agents on open weights this year.
The task and the numbers
Same prompt on both, same repo context, same agent harness. The job wasn't a toy benchmark. It was the kind of thing I actually do in a working week: take a half-formed project plan, restructure it, and start writing the code. Here's what each run looked like.
Model
