tokens-per-dollar · prices read

Codex credits to tokens

Codex credits are charged per token. On GPT-6.1 Sol one credit buys 20,000 input tokens, 400,000 cached input tokens or 4,000 output tokens (50, 2.5 and 250 credits per million). OpenAI does not state a dollar price per credit for Plus or Pro; its own figures put it at $0.04: $100 of workspace credits is 2,500 credits, and its Enterprise USD rate card is 25 credits to the dollar. At that price every rate in the table below comes to the model's API list price. What your own month comes to is a sum over your tokens: npx --allow-git=root github:agentwares/tokens-per-dollar does it from the session files Codex keeps.

The rate card, per model

ModelCredits per million: input · cached · outputTokens per credit: input · cached · outputPer million at $0.04: input · output
GPT-6 Astra250 · 25 · 1,2504,000 · 40,000 · 800$10.00 · $50.00
GPT-6.1 Sol50 · 2.5 · 25020,000 · 400,000 · 4,000$2.00 · $10.00
GPT-6 Sol50 · 5 · 25020,000 · 200,000 · 4,000$2.00 · $10.00
GPT-6 Luna2.5 · 0.25 · 12.5400,000 · 4,000,000 · 80,000$0.10 · $0.50
GPT-5.6 Sol100 · 10 · 50010,000 · 100,000 · 2,000$4.00 · $20.00
GPT-5.6 Terra50 · 5 · 30020,000 · 200,000 · 3,333$2.00 · $12.00
GPT-5.6 Luna5 · 0.5 · 30200,000 · 2,000,000 · 33,333$0.20 · $1.20
GPT-5.5125 · 12.5 · 7508,000 · 80,000 · 1,333$5.00 · $30.00
GPT-5.3-Codex43.75 · 4.375 · 35022,857 · 228,571 · 2,857$1.75 · $14.00
GPT-5.243.75 · 4.375 · 35022,857 · 228,571 · 2,857$1.75 · $14.00

Standard speed. Codex credit billing has no separate cache-write charge: tokens written to the cache count as input. Fast mode, where available, uses twice the Standard rate on purchased credits. GPT-5.6 Sol's rates are promotional, at least through 21 November 2026. The dollar column is derived (OpenAI does not state a price per credit for Plus or Pro; $0.04 is derived from its own figures: $100 of workspace credits is 2,500 credits, and its Enterprise USD rate card is 25 credits to the dollar on every row).

Credits and the plans

ChatGPT Plus ($20 a month) and Pro (from $100: tiers at $100, $200 and $500 a month, billed monthly) include Codex. Plus has a five-hour window and a weekly limit; Pro has no five-hour limit. OpenAI publishes local-message estimates per five hours for Plus, which it says are not fixed limits, and no allowance for either plan in tokens or credits. Plus and Pro users who reach the limit can buy credits and keep working at this rate card.

What OpenAI does not publish, Codex records: it writes the plan and the percentage of the 5-hour and weekly windows used into its session files. So for the plan you were on, whether it carried your month is measured, not estimated.

$ npx --allow-git=root github:agentwares/tokens-per-dollar demo
Codex: 628 requests, 25.8M tokens, $34.68 at OpenAI API list prices
  busiest 5 hours $2.41 · busiest week $9.16 (API list prices)
  Codex's own readings (Plus): busiest 5-hour window 48%, busiest week 12% of the limit, never at it

  plan           fee/mo  a month like this            how
  ChatGPT Plus      $20             $20.00  measured  Codex's readings: never at the limit
  Copilot Pro       $10             $29.68  estimate  $34.68 of credits, $15.00 included ²
  Codex credits       —             $33.50  priced    837 credits at $0.04 ³
  OpenAI API          —             $34.68  priced    list prices × your tokens
  Copilot Pro+      $39             $39.00  estimate  $34.68 of credits, $70.00 included ²
  Copilot Max      $100            $100.00  estimate  $34.68 of credits, $200.00 included ²
  ChatGPT Pro      $100       from $100.00            no 5-hour limit; weekly not published

  Cheapest priced: ChatGPT Plus, $20.00 a month (measured), $9.68 less than Copilot Pro ($29.68).
Synthetic history: the real output of npx --allow-git=root github:agentwares/tokens-per-dollar demo on generated transcripts, not anyone's usage (as of 7 October 2026).

Changelog: what the vendors changed, and when

Dates are the vendors' own. “As reported” means the press quoting the vendor, because the vendor's own post could not be read. Each entry says what changed, never why.

Run it on your own logs

On the machine where you use Claude Code or Codex:

$ npx --allow-git=root github:agentwares/tokens-per-dollar

It reads the transcripts Claude Code and Codex keep on your machine, takes the token counts, and prices your last 30 days on every plan on this page. Free and MIT; it makes no network request and sends nothing. --json prints every price with its source and the date it was read.

For a team: seatledger keeps everyone's coding-agent ledger in one place and flags when the token rate moves. Free for up to 3 developers; email alerts on Team, $49 a month.

Sources