The cost comparison is the right angle. What I've found since then is that you don't need to pick one or the other. I've been running a setup where Claude Code routes through a proxy serving GLM-4.7 (successor to 4.6) alongside DeepSeek and Kimi K2.5, all under a single $30/mo subscription https://reading.sh/how-to-get-3x-claude-rate-limits-for-30-a-month-1d3fdb8658df
You get 3x the rate limits of Claude Pro and access to the full model library. The junior vs senior framing is spot on though. I use the cheaper models for routine work and swap to heavier ones when the task needs it.
Don't Forget the GLM coding Plans which is 3$ a Month without any limits.
And the limits that exists are time based or promot based and not Token Based.
So even a basic GLM 4.6 plan is 3X better then a claude code basic plan.
The cost comparison is the right angle. What I've found since then is that you don't need to pick one or the other. I've been running a setup where Claude Code routes through a proxy serving GLM-4.7 (successor to 4.6) alongside DeepSeek and Kimi K2.5, all under a single $30/mo subscription https://reading.sh/how-to-get-3x-claude-rate-limits-for-30-a-month-1d3fdb8658df
You get 3x the rate limits of Claude Pro and access to the full model library. The junior vs senior framing is spot on though. I use the cheaper models for routine work and swap to heavier ones when the task needs it.
Worth to give it a try definitely!