Here's a detailed breakdown of the **limits for the GLM Coding Lite plan**, sourced from the official Z.AI docs and pricing guides (data current as of July 2026): --- ## 📊 GLM Coding Lite — Limits at a Glance | Metric | Limit | |---|---| | **Prompts per 5-hour rolling window** | **~80 prompts** | | **Weekly cap** (7-day rolling) | **~400 prompts** | | **MCP monthly quota** (web search + web reader calls) | **100 calls/month** | | **Vision understanding** | Uses the 5-hour prompt pool (no separate cap) | --- ## 💰 Pricing - **Standard:** $18/month - **Promo (30% off):** $12.60/month (through Sept 2026) - **Yearly:** $151.20/year --- ## 🧠 How Quota Consumption Works It's not a simple 1:1 token count — the system uses **prompt-based limits**, not token-based limits. Here's what that means: 1. **1 prompt = 1 user turn** in your coding tool (e.g., one chat message or one agent action). Each prompt typically allows **15–20 model calls** internally — so you get **tens of billions of tokens per month** at ~1% of standard API pricing. 2. **Model usage multipliers (important!):** | Model | Peak hours | Off-peak hours (until Sept 2026) | |---|---|---| | **GLM-5.2 / GLM-5-Turbo** | 3× quota consumption | 1× (promo) | | **GLM-4.7 / GLM-4.5-Air** | 1× | 1× | ⚠️ So if you use GLM-5.2 during peak hours, a single prompt can consume **3 of your 80 prompts**. 3. **No overage billing** — Once you hit the cap, the plan stops until the next 5-hour cycle refreshes. It does **not** auto-charge your account balance. 4. **Peak vs. off-peak:** Peak hours consume quota faster, but Z.AI's limited-time promotion through September 2026 makes off-peak GLM-5.2 usage only **1×**. --- ## 🔄 Refresh Cycles - **5-hour rolling window:** ~80 prompts, then you wait for refresh - **Weekly cap (~400 prompts):** resets on a 7-day cycle from when you ordered - **MCP calls (100/month):** monthly reset --- ## 🛠 Supported Tools The Lite plan works in all the same tools as higher tiers — Claude Code, Cursor, Cline, Roo Code, OpenClaw, Continue.dev, and 20+ other supported coding assistants. --- **Bottom line:** The Lite plan gives you roughly **80 requests per 5 hours** (or ~16 prompts/hour sustained), with a **weekly ceiling of ~400**. If you use the flagship GLM-5.2 model during peak times, you'll burn through that ~3× faster. For light, casual coding assistance this is fine — for heavy daily agentic work, you'd want Pro or Max.