> I suspect Claude might be faster and therefore cheaper, but maybe not by a lot.
While Jarred used Mythos-class model, some open weights, if they were as capable (certainly, GLM 5.2 looks the part), would have been way, way cheaper than professionals.
Approx costs:
DeepSeek v4 Pro & Mimo v2.5 Pro $3,426 ($2,567 / $600 / $259)
Tencent HY3 $3,892 ($1,180 / $552 / $2,160)
GLM 5.2 $30,016 ($8,260 / $3,036 / $18,720)
Qwen 3.7 Max $37,925 ($14,750 / $5,175 / $18,000)
Claude Opus 4.8 & GPT 5.5 xhigh $82,750 ($29,500 / $17,250 / $36,000)
5.9 billion uncached input tokens, 690 million output tokens, 72 billion cached input token reads.
I did some more sizeable work with GLM 5.2 on Max reasoning (planning and implementing 8 features end to end) and it performed pretty well, but worse than Opus 4.8, with largely the same adversarial agent review loop.
Opus 4.8 still found some real issues afterwards and spent about an hour fixing things, before the code was good enough to ship. Overall promising, wrote about it here: https://blog.kronis.dev/blog/z-ai-s-glm-5-2-is-a-great-model...
The GLM Coding Plan seems to have lower token limits than the corresponding Anthropic Max subscriptions, but if you had to pay API rates for some LLM to do work somewhat reliably, it's a no brainer (unless you're swimming in money that you can give away and value your time more).
But not all models are equally capable, so I don't know your basis of comparison is even valid, let alone the numbers.