NeverMetered vs ChatGPT and Codex: plans, limits and fit

Updated 10 November 2026

Codex isn't sold on its own. You get it as part of a ChatGPT plan, and your Codex usage is metered against that plan. NeverMetered is a different shape of product: one open-weight model with no usage meter, priced by how many requests run at once. This page puts the two side by side so you can tell which problem each one solves.

How Codex is priced and limited

Per OpenAI's pricing page on 10 November 2026: Free ($0) and Go ($8 a month) list Codex for quick, lightweight tasks with no published numbers. Plus ($20 a month) covers Codex on the web, CLI, IDE extension and iOS, and runs on a rolling 5-hour window plus a weekly limit that may apply. Pro is "from $100 a month", with $100, $200 and $500 tiers, and OpenAI says Pro currently has no 5-hour window though weekly limits may apply. OpenAI doesn't publish the multiple each Pro tier gives, so we don't either.

OpenAI's estimates for Plus are ranges per 5-hour window, and they depend heavily on the model: from 5 to 45 local messages on GPT-6 Astra up to 350 to 3,000 on the lighter GPT-6 Luna. OpenAI is clear these are estimates, not fixed numbers. The full breakdown is in our Codex usage limits guide.

How NeverMetered is priced and limited

NeverMetered has no token quota, 5-hour window or weekly cap on paid plans. You pay for parallel streams: Solo is $24 a month for 1 stream and Pro is $67 a month for 3 streams. Solo allows 256K context and up to 15 minutes per request; Pro allows 256K context and up to 15 minutes per request (see pricing). Requests beyond your streams wait briefly in a queue rather than failing. It runs one model, Qwen3.8-27B, as a 4/5-bit quantized build, and its own how it runs page states that it's slower than frontier APIs and that its quality evals are in progress.

ChatGPT plans with CodexNeverMetered
PriceGo $8, Plus $20, Pro from $100 ($100, $200 or $500)Solo $24 or Pro $67 per month
ModelsOpenAI's GPT-6 family, chosen per taskQwen3.8-27B only
Limits5-hour window plus weekly limit on Plus; Pro currently weekly onlyHow many requests run at once
Published allowanceRanges per window, described as estimatesNone: there is no token allowance to run out of
ResetsRolling window; paid resets offered to some accountsNothing to reset
ToolsCodex CLI, IDE extension, web, iOSCodex CLI, Claude Code, OpenCode, Cursor and the other supported tools, over an OpenAI- and Anthropic-compatible API
Speed and strengthFrontier modelsA smaller open-weight model, slower, strong for its size

Where Codex is the better choice

  • You want OpenAI's strongest models for hard tasks, cloud tasks and code review, and you already pay for ChatGPT.
  • Your usage fits inside a Plus window most days, so the limits never get in the way.
  • You want the polished cloud and IDE experience around Codex, not just a model endpoint.

Where NeverMetered is the better choice

  • Your weekly limit runs out before your 5-hour window ever fills, which happens when a few large agent runs read and re-read a lot of files.
  • You want long unattended runs without watching a meter.
  • You'd rather have the same bill every month, whatever you run.
  • You use Codex CLI already: pointing it at NeverMetered is a config change. See the Codex CLI setup guide.

Who should pick which

If you are...Pick
Using Codex a few times a week for focused tasksStay on your ChatGPT plan
Hitting the weekly Codex limit regularly on PlusKeep Plus for the hard tasks, send routine volume to NeverMetered
Considering Pro mainly for more headroomCheck how much of your usage is routine first: a flat-rate lane may cover it for less
Needing a model OpenAI doesn't shipNeverMetered or another provider; Codex is tied to OpenAI's models on a ChatGPT sign-in
Is there a flat-rate alternative to Codex's usage limits?
Yes. NeverMetered charges by parallel streams, so there's no 5-hour window or weekly cap, and extra requests queue briefly rather than failing. It runs one open-weight model, not OpenAI's, so it works best as a second lane next to a ChatGPT plan.
Can I use Codex CLI with NeverMetered?
Yes. Codex CLI can point at an OpenAI-compatible endpoint, and NeverMetered is one. The setup guide has the config.
Does ChatGPT Pro remove Codex's limits?
Per OpenAI's pricing page, Pro currently has no 5-hour window, but weekly limits may still apply, and OpenAI's wording is "currently", so it may change. OpenAI hasn't published a weekly figure for any plan.
Is NeverMetered as good as GPT-6 models at coding?
No, and it doesn't claim to be. It's a capable open-weight model for its size and suits routine work. For your hardest problems a frontier model is the better tool.
What happens when I hit my Codex limit?
Codex can finish the turn it's on, subject to fair use, then the next turn is blocked until the window resets, you use a saved reset or you buy credits. The limits guide covers each option.