Model Credit Costs
Not every message costs the same number of credits — cost scales with how expensive the model is to run.
| Cost per message | Models |
|---|---|
| 1 credit | Gemini 3.1 Flash-Lite, Claude Haiku 4.5, ChatGPT 5.4 Mini, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite, Gemini 3 Flash (Preview) |
| 2 credits | Gemini 3.7 Flash, Gemini 3.1 Pro (Preview), Gemini 2.5 Pro, ChatGPT 5.4 Auto |
| 3 credits | Claude Sonnet 5, Claude 4.6 Sonnet, Claude 4.5 Sonnet, Claude 3.7 Sonnet, ChatGPT 5.4 Reasoning |
| 6 credits | ChatGPT 5.5 |
| None | Any model running on your own key (Max) |
ChatGPT 5.5 has a band to itself because it is the only model here priced above the Sonnet tier — three credits would not cover what one of its requests actually costs to serve.
Full-Page Deep Scan surcharge
Turning on Deep Scan adds a surcharge on top of the model's normal cost, since it processes far more text per request:
The surcharge is the model's own cost doubled, plus one.
| Normal cost | With Deep Scan |
|---|---|
| 1 credit | 3 credits |
| 2 credits | 5 credits |
| 3 credits | 7 credits |
| 6 credits | 13 credits |
Deep Scan is a Pro feature, so this table never applies to a guest or a Free account — which is why ten and fifteen credits a day stay ten and fifteen messages a day on those plans.
If you're trying to stretch a monthly credit allowance further, everyday questions on a 1-credit model go a lot further than a Deep Scan request on ChatGPT 5.5 (13 credits) — save the heavier combination for when you actually need it.