GLM 5.2 with ClinePass: 61M tokens in one 5-hour coding session
so recently i bought clinepass to give it a try. I used GLM 5.2 as the main workhorse and got solid usage for the price (got it at $1.99).
I use GPT 5.5 in codex as the planner and reviewer. it writes the proposal and execution plan, then GLM 5.2 acts as the implementer.
the result was quite solid. I got two infra refactors done in a single 5-hour session. it consumed around 61M tokens for that one session. the session itself used around 20% of the monthly limit. I started at 3%, based on the image. so if we estimate from that, the monthly limit would be gone in about 5-6 sessions.
well, I can’t complain since it only cost $1.99. it’s good considering the price.
as for GLM 5.2 itself, it literally blew my mind, lol. solid executor. I had codex review the code generated by GLM, and it couldn’t find any major issues. everything aligned with the plan codex gave.
so I think this is a pretty comfortable workflow for people looking for an affordable setup: a $20 codex subscription for planning, review, and the hardest tasks, then a cheaper open-model subscription like cline or opencode for execution.
btw, I tested it on my side project: an AI token usage tracker. it’s local-first and already supports most coding agents like codex, claude code, opencode, antigravity, cline, and zcode.
if you want to give it a try:
https://github.com/fikrilal/burnly