OpenAI Codex AI credits, explained
OpenAI Codex meters usage in credits. One credit costs $0.04, and every model's credit rate card equals its direct OpenAI API dollar rate card multiplied by 25, so there is no per-model markup: Codex credits are API list pricing in a different unit. Codex plans meter local messages and cloud chats in a shared 5-hour window plus weekly limits, and OpenAI publishes no fixed monthly credit allotment per plan. The pages below explain the rate card per model, and explain what OpenAI has not published.
- Unit
- 1 credit = $0.04
- Markup over the API
- 0%, by construction
- Verified
- 2026-08-22
- Edition
- 2026-08.3
Per-model rate card: credits, dollars, and the API
Credits per 1 million tokens as OpenAI meters them, the dollar value of those credits at $0.04 each, and OpenAI's direct API dollar rate for the same model. The last two columns match on every model, which is the proof that Codex credits carry zero markup over the API.
| Model | Category | Credits / 1M (in / out) | $ equivalent / 1M (in / out) | Direct API / 1M (in / out) |
|---|---|---|---|---|
| GPT-5.6 Sol | Powerful | 100 / 500 | $4.00 / $20.00 | $4.00 / $20.00 |
| Daybreak Bluegated | Cyber | 100 / 500 | $4.00 / $20.00 | $4.00 / $20.00 |
| Daybreak Redgated | Cyber | 312.5 / 1,875 | $12.50 / $75.00 | $12.50 / $75.00 |
| GPT-5.6 Terra | Versatile | 50 / 300 | $2.00 / $12.00 | $2.00 / $12.00 |
| GPT-5.6 Luna | Lightweight | 5 / 30 | $0.20 / $1.20 | $0.20 / $1.20 |
| GPT-5.4 mini | Lightweight | 18.75 / 113 | $0.75 / $4.52 | $0.75 / $4.50 |
| GPT-5.5prior gen | Powerful | 125 / 750 | $5.00 / $30.00 | $5.00 / $30.00 |
| GPT-5.4prior gen | Versatile | 62.5 / 375 | $2.50 / $15.00 | $2.50 / $15.00 |
Plans and how credits meter
| Plan | Price | Audience | Metering |
|---|---|---|---|
| ChatGPT Free | $0.00/mo | Individual | five hour window |
| ChatGPT Go | $8.00/mo | Individual | five hour window |
| ChatGPT Plus | $20.00/mo | Individual | five hour window |
| ChatGPT Pro | $100.00/mo and up | Individual | five hour window |
| ChatGPT Business | $20.00/user/mo | Org | flexible |
| ChatGPT Enterprise | custom | Org | flexible |
Local messages and cloud chats share a 5-hour window, with additional weekly limits. Paid plans can purchase extra credits to keep working once the window or weekly limit is hit. OpenAI publishes no fixed monthly credit allotment for any plan, so this page does not invent one.
Plans and metering, in full - How many credits a task actually costs
Credits per model
How many credits each current-generation model burns per request, ranked cheapest first. See all models.
How much a credit balance is worth
Round credit amounts converted to dollars and per-model token equivalents.
Compare
Model-vs-model, plus how Codex credits line up against Copilot's AI credits and the raw OpenAI API.
GPT-5.6 Sol vs GPT-5.6 Terra,GPT-5.6 Terra vs GPT-5.6 Luna,GPT-5.6 Sol vs GPT-5.6 Luna,Codex credits vs GitHub Copilot AI credits,Codex credits vs the OpenAI API price.
Dated edition
Edition 2026-08.3, last modified 2026-08-22. Second recorded repricing. GPT-5.6 Sol fell from 125/12.5/750 to 100/10/500 credits ($5/$0.50/$30 to $4/$0.40/$20), and Daybreak Blue moved with it because OpenAI pins it to Sol credit rates. Input-side rates fell 20%, output-side 33%, and 1 credit = $0.04 held exactly. The cut landed between the 2026-08-19 and 2026-08-22 checks; on the API dollar card the ai-price-index sentinel brackets it to between 2026-08-21 07:47 UTC and 2026-08-22 02:12 UTC. OpenAI calls the new rate promotional and available at least through 2026-11-21, which this schema cannot express as a scheduled reversion. Also lowers the stated per-message credit range from 5-40 to 5-30. Sol is now cheaper per token than prior-generation GPT-5.5, which still sits at the old 125/12.5/750.
Use this data
The full rate table is available as machine-readable rates.json and rates.csv, verified 2026-08-22 against OpenAI's published rate card. Free to reuse with attribution (CC BY 4.0).