What This Is
A Juejin author's 5-month bill tells us: heavy personal AI users are now keeping monthly costs under 100 yuan. His main work is daily content for WeChat Official Accounts and Xiaohongshu, with side dev projects on top—AI is embedded across research, drafting, and code generation.
The bill shows DeepSeek API spending went from ~55 yuan in February, down to 20 yuan in April, then back up to 60 yuan in July. Combined with Trae international version (~50 yuan) and Jimeng AI (69 yuan) subscriptions expiring and being dropped, monthly costs fell from ~200 yuan to ~100 yuan. He also built a Chinese LLM pricing comparison table, with the core logic of calculating the breakeven between "subscription price ÷ quota" and the API pay-per-use unit cost: below the breakeven, pay-per-use wins; above it, a subscription wins.
Industry View
This bill is valuable because the triggering events are representative: the DeepSeek V3-to-V4 price hike and Trae international subscription expiry landed simultaneously, forcing a tool-stack reshuffle—Trae replaced with OpenCode, Jimeng swapped for free quotas plus Doubao/Qwen alternatives, and the IDE moved to NeoVim with a terminal-first workflow.
But a warning: the "100 yuan monthly fee" is an extreme optimization result, built on stacking 200 free requests per day, off-peak API usage, and templating repeated workflows to cut Token (the smallest unit LLMs charge by, roughly mapped to text length) consumption—difficult for average users to replicate.
Another overlooked signal: the author notes a DevOps colleague looking at AWS bills quipped, "costs creep back up if you don't check for a few months." AI tool cost governance is an ongoing discipline, not a one-time decision.
Impact on Regular People
For individual professionals: AI tool pricing has split into "pay-per-use" and "subscription" tracks—estimate your monthly call volume first before deciding, and don't get locked into subscriptions chasing free-tier hooks.
For enterprise IT: the playbook of cutting subscriptions and swapping in free tools from individual budgets largely doesn't apply—compliance, data security, and collaboration permissions carry hidden costs that dwarf subscription savings.
For the consumer market: Chinese LLM price hikes are now industry consensus—the "freebie window" is narrowing, and pay-per-use will become the mainstream option.