ChatGPT / Codex Plus

OpenAI · Coding agent plan · openai/plus

Price a month

$20

Est. $/Mtok at full use

$0.089

Checked

5 Oct 2026

Models included

• gpt-6.1-sol
• gpt-6-sol
• gpt-6-luna
• gpt-6-astra
• gpt-5.6-*

Published limits

• “GPT-6.1 Sol | 15-160 [local messages per five-hour period, Plus]”

  5-hour rolling window · 160 messages

• “GPT-6 Astra | 5-45”

  5-hour rolling window · 45 messages

• “GPT-6 Luna | 350-3,000”

  5-hour rolling window · 3,000 messages

• “Local messages and cloud chats share your plan's usage allowance. Weekly limits may also apply.”

  weekly cap · unknown

Fair use

“If you reach your usage limits during an active turn, the agent will be able to continue working on that turn, subject to fair use limits.”

Token math, step by step

The assumption: one request is 20,000 input tokens (18,000 of them cache reads) plus 1,500 output = 21,500 tokens, and a month of full use is 4 usable 5-hour windows a day × 30 days = 120 windows.

This estimate does not follow from the caps alone (credit pool, dollar pool or a measured conversion). The dataset’s method, verbatim:

“STANDARD: 1 request/prompt/message = 1 agentic model call of 20,000 input tokens (18,000 cache-read + 2,000 uncached) + 1,500 output = 21,500 tokens. Window plans: 4 usable 5h windows/day x 30 days = 120 windows/month, then capped by weekly cap x 30/7 and by any monthly cap (whichever binds first). Dollar-pool plans: pool / blended reference rate for the same token mix at Claude Sonnet 5.5 list ($2 in / $0.20 cache-read / $10 out per Mtok) = $1.0512/Mtok. Codex: midpoint of published GPT-6.1 Sol range (15-160 -> 87.5 local messages/5h) x 120 windows; weekly cap unpublished so this is an UPPER bound on the window side, but a Codex 'local message' is a full agent turn (many model calls) so 21.5k tokens/message is a conservative token count.”

1. Tokens a month at full use: 225,750,000.
2. USD price ÷ (tokens ÷ 1,000,000) = $20 ÷ 225.75 = $0.089 per 1M.
3. The same mix through the Claude Sonnet 5.5 API costs $1.05 per 1M, so at full use this plan is 12× cheaper.

“At full use” means using every window to the cap. Real use is usually well below it.

Sources

• Source: learn.chatgpt.com
• Checked 5 Oct 2026; last verified 5 Oct 2026.
• Confidence: vendor page.

Price history

One snapshot so far, from 5 Oct 2026. Price history is planned.

Back to ComparePrice alerts

Price-drop alerts are planned; the waitlist for them is open.