ChatGPT, Claude and Gemini all meter usage now — none tells you what one message costs

Ok_Negotiation_2587 · reddit · 2026-11-05

Having read all three vendors' help pages, the author maps their convergence on opaque metered allowances:

ChatGPT: from December, Plus/Pro/Business allowances are token-based and shared between ChatGPT and Codex; instant chat is free; usage visible in Settings.
Claude: paid plans get a 5-hour session limit plus weekly caps, driven by conversation length, complexity, model and effort level; Claude.ai and Claude Code share the pool, and Anthropic may adjust limits "at our discretion."
Gemini: "compute-based" limits refreshing every 5 hours with weekly caps; AI Plus is "2x standard," AI Pro "4x" — but the standard is never published, only a relative cost table from "Very low" to "High."

The core critique: you see a percentage of an unstated total, learn a task's cost only after running it, and "4x of an undefined number" can't be compared across vendors or planned around. The author gets why (wildly varying serving costs, capacity management), but the uncertainty lands on paying users whose work now depends on these tools.

He also shares a DIY method: note the percentage, run a typical task, note again; compare fresh vs. long chats to build your own numbers. He promotes his free extension AI Toolbox, which surfaces Gemini's and Claude's usage figures and reset times under the message box — though it can't predict per-message cost either.

Original post →

ChatGPT, Claude and Gemini all meter usage now — none tells you what one message costs

More from Models