progressEdd's llms, ai, and dev corner

ProgressEdd, a developer, criticized GLM's updated coding plan for reduced token allowances, stating he will not upgrade and will keep his Legacy V2 plan due to GLM's openness in publishing weights and training techniques compared to Anthropic. He plans to save for hardware to run Deepseek V4 Flash locally and build datasets from his pi sessions. If his plan expires, he will follow opencode Reddit's advice to use OpenCode Go with MiniMax, recommending GLM5.1 as planner/reviewer and MiniMax M2.7 as a workhorse to avoid limits.

Update, GLM altered the deal, if you sign up now, the coding plan isn’t as good anymore especially If you are a heavy consumer of tokens. I am unhappy by this decision and do not plan on upgrading. I will continue to hold my Legacy V2 plan because they more open in publishing their weights and training techniques than Anthropic. At this point, I’m saving up and waiting until hardware is cheap enough to run Deepseek V4 Flash locally like sentdex Also until hardware gets cheaper, I think I’ll try to build out the datasets based on my pi sessions for my own data flywheel like wendell If my plan does expire, I will probably follow the guidance from the opencode reddit which recommends open code go + minimax I ended the OC GO limits in 2 full 5 hours sessions of GLM5.1+Qwen3.6Plus usage to implement one complex task. Speed was pretty good. If you go mainly for Qwen3.6 Plus and MiniMax M2.7, you’d have much more usage. Use the GLM5.1 as planner and reviewer only. Consider buy MiniMax Coding Plan + OC Go and use M2.7 as a workhorse, then you’d never hit the limits and M2.7 is fast even on 10$ plan.