DALL-E Pricing 2026
DALL-E 3 remains accessible, but OpenAI marks its API as deprecated. New buyers should compare the current ChatGPT Images experience with supported OpenAI image API models rather than treating old DALL-E prices as the current product.
Affiliate link, we may earn a commission at no extra cost to you.
Before you pay full price for DALL-E / ChatGPT Images, check for a deal
AIDeals.app tracks verified discounts on the AI tools you already want, annual-billing savings, student and nonprofit programs, free tiers, and promo codes, every deal checked by hand and stamped with its verification date. Plus free AI cost calculators to plan your stack.
AIDeals.app is our sister site for verified AI-tool discounts.
Updated for the current OpenAI image stack
DALL-E pricing in 2026: the legacy keyword and the current buying decision
DALL-E 3 API access is deprecated. ChatGPT still offers a DALL-E GPT, while the main ChatGPT Images experience and supported developer image models have moved on.
A useful pricing answer must separate three products that old comparison pages collapse together: the DALL-E GPT inside ChatGPT, the current ChatGPT Images experience, and image generation through the OpenAI API. They do not share one simple per-image price.
OpenAI says DALL-E 3 API access remains available but is deprecated and slated for removal. I would not begin a new product integration around it. Existing users should inventory model names, image sizes, quality settings and tests, then plan migration using the current image generation documentation.
For casual creation, evaluate ChatGPT as a subscription
ChatGPT Images is available across tiers, with access and limits varying by plan. The value includes conversational editing and other ChatGPT capabilities, so dividing the subscription by an assumed image quota creates a misleading unit price.
For applications, price the supported API model
API cost depends on the selected model, quality, dimensions and input or output tokens. Use the live model page and pricing calculator at purchase time rather than a DALL-E 3 table copied from 2023.
Include iteration in the budget
A usable asset may require several generations, edits and variations. Estimate cost per accepted asset from a pilot, not cost per first output.
Separate generation from production cost
Storage, moderation, retries, queues, review, design finishing and delivery can exceed model spend. A complete budget includes the workflow around the image.
Do not promise fixed ChatGPT quotas
Usage limits can change with plan and system conditions. Link to the current plan and product documentation instead of claiming a permanent number of images per day.
Treat legacy access as a migration case
If DALL-E 3 currently meets a production need, capture representative prompts and acceptance tests. Compare outputs before switching rather than assuming the newer model behaves identically.
How I would choose and budget the route
- 1
Name the job
Do: Decide whether a person is creating interactively or software is generating programmatically.
Example: A marketer editing campaign concepts has a different cost model from an app producing thumbnails.
Checkpoint: The interface requirement is clear.
- 2
Build a representative prompt set
Do: Collect twenty real tasks including text, editing, transparent backgrounds and difficult compositions.
Example: Do not benchmark only the easiest hero image.
Checkpoint: Success criteria are written before testing.
- 3
Record full consumption
Do: Track attempts, quality, dimensions, edits, latency and accepted results.
Example: Five low-cost failed attempts may cost more operationally than one higher-quality output.
Checkpoint: You know cost per accepted asset.
- 4
Check current availability
Do: Confirm supported model, endpoint, rate limits and plan terms in OpenAI's official documentation.
Example: Do this again before publication or procurement.
Checkpoint: No decision relies on an old DALL-E table.
- 5
Add workflow cost
Do: Estimate moderation, human review, design correction, storage and failed-job handling.
Example: Text-heavy images may need manual QA even when prompt adherence improves.
Checkpoint: The budget covers production.
- 6
Plan migration
Do: For legacy DALL-E API use, create compatibility tests and a rollback path.
Example: Compare aspect ratio, edit behaviour, text rendering and safety responses.
Checkpoint: Removal of the deprecated model is survivable.
Pricing claims I would not publish
DALL-E is the default ChatGPT image model
OpenAI distinguishes the current ChatGPT Images experience from the still-accessible DALL-E GPT.
ChatGPT Plus includes a fixed daily quota
Do not turn variable limits into a permanent promise.
The API costs one price per image
Model, quality, size and token use affect cost.
Generation cost equals asset cost
Iteration and human finishing belong in the calculation.
Deprecated means already unavailable
It remains accessible, but is slated for removal; that distinction matters to existing integrations.
Commercial use needs no review
Teams should check current terms, policies, trademarks, likeness rights and their own risk requirements.
Prompts for a defensible image-cost test
Benchmark designer
Turn these real image requests into a benchmark covering composition, embedded text, brand constraints, editing, transparency, aspect ratios and safety-sensitive edge cases. Define pass criteria without changing the requests.
Run the same set across candidate routes.
Cost calculator
Using the official rates I paste, calculate cost per attempt and per accepted asset for low, medium and high acceptance scenarios. Include retries, edits and review time. Show formulas and do not invent quotas or prices.
Paste current official pricing.
Migration test
Create a compatibility checklist for moving from DALL-E 3 to the supported image model. Cover request parameters, sizes, quality, edits, output handling, safety responses, latency and visual acceptance tests.
For existing API users.
Procurement summary
Summarise the choice between interactive ChatGPT Images and API generation for this team. Compare control, integration, collaboration, variable usage, governance and total workflow cost. Mark every fact that needs current vendor verification.
Useful before plan approval.
Reader workbook
Working notes for dall-e pricing in 2026: the legacy keyword and the current buying decision
Reading the guide is only the first pass. The notes below turn its recommendations into evidence you can inspect, discuss with another person, and revise. Complete them with real material from your situation. Do not let an AI assistant fill gaps with plausible facts. When a policy, price, specification, source, system state, or personal experience matters, open the authoritative record and put the verified detail in your working document.
Working note 1: For casual creation, evaluate ChatGPT as a subscription
Begin with a concrete example from the last thirty days. Record what happened, what information was available at the time, who made the decision, and what the result was. Then apply the principle above to that example. The useful output is not a general agreement that the principle sounds sensible; it is one changed action, one piece of evidence you will collect, and one condition that would make you choose a different approach. Write those three items in language another person could audit.
Evidence to leave behind
A dated note that connects “For casual creation, evaluate ChatGPT as a subscription” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
Working note 2: For applications, price the supported API model
Test this principle against a difficult case rather than the easiest one. List the constraint most likely to be ignored, the person who carries the downside if the advice is wrong, and the source that can settle a factual disagreement. Next, describe a small trial that is reversible and produces a visible result. Decide in advance what would count as improvement, no change, or harm. This turns a broad recommendation into a decision with limits instead of another optimistic intention.
Evidence to leave behind
A dated note that connects “For applications, price the supported API model” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
Working note 3: Include iteration in the budget
Explain this idea to a colleague, teacher, adviser, reviewer, or teammate without using jargon. Ask them where the explanation assumes knowledge that has not been demonstrated. Add the missing source, example, calculation, test, or observation. Finally, write the strongest reasonable objection and a response that acknowledges the trade-off. If the response depends on a vendor claim or an AI answer, mark it unverified until it has been checked against a primary source or real result.
Evidence to leave behind
A dated note that connects “Include iteration in the budget” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
Working note 4: Separate generation from production cost
Begin with a concrete example from the last thirty days. Record what happened, what information was available at the time, who made the decision, and what the result was. Then apply the principle above to that example. The useful output is not a general agreement that the principle sounds sensible; it is one changed action, one piece of evidence you will collect, and one condition that would make you choose a different approach. Write those three items in language another person could audit.
Evidence to leave behind
A dated note that connects “Separate generation from production cost” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
Working note 5: Do not promise fixed ChatGPT quotas
Test this principle against a difficult case rather than the easiest one. List the constraint most likely to be ignored, the person who carries the downside if the advice is wrong, and the source that can settle a factual disagreement. Next, describe a small trial that is reversible and produces a visible result. Decide in advance what would count as improvement, no change, or harm. This turns a broad recommendation into a decision with limits instead of another optimistic intention.
Evidence to leave behind
A dated note that connects “Do not promise fixed ChatGPT quotas” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
Working note 6: Treat legacy access as a migration case
Explain this idea to a colleague, teacher, adviser, reviewer, or teammate without using jargon. Ask them where the explanation assumes knowledge that has not been demonstrated. Add the missing source, example, calculation, test, or observation. Finally, write the strongest reasonable objection and a response that acknowledges the trade-off. If the response depends on a vendor claim or an AI answer, mark it unverified until it has been checked against a primary source or real result.
Evidence to leave behind
A dated note that connects “Treat legacy access as a migration case” to one actual decision, names the evidence used, records uncertainty, and identifies the next person or check required before the decision becomes final.
A review record for how i would choose and budget the route
Keep one row for every pass through the workflow. The record should make progress and failure equally easy to see. A polished output with no trace of its sources, assumptions, checks, or human decisions is difficult to improve and dangerous to trust. The following review questions are deliberately tied to the steps above.
After step 1
Review “Name the job”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: The interface requirement is clear. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
After step 2
Review “Build a representative prompt set”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: Success criteria are written before testing. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
After step 3
Review “Record full consumption”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: You know cost per accepted asset. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
After step 4
Review “Check current availability”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: No decision relies on an old DALL-E table. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
After step 5
Review “Add workflow cost”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: The budget covers production. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
After step 6
Review “Plan migration”
Record what you actually did, not what the plan said you would do. Attach the relevant output or source. Use this checkpoint as the acceptance test: Removal of the deprecated model is survivable. If it is not met, note whether the cause was missing information, weak skill, unclear ownership, insufficient time, a faulty assumption, or an external constraint. Choose one correction and repeat the smallest affected step rather than restarting the entire workflow.
A red-team pass before you rely on the result
Use the failure modes from this guide as a final challenge, not as a warning box you read and forget. Assign each one to a reviewer, or take them one at a time yourself. The reviewer should point to evidence in the work and should be allowed to say that the evidence is insufficient.
Could “DALL-E is the default ChatGPT image model” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
Could “ChatGPT Plus includes a fixed daily quota” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
Could “The API costs one price per image” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
Could “Generation cost equals asset cost” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
Could “Deprecated means already unavailable” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
Could “Commercial use needs no review” be happening here?
Find the strongest sign that it is, then the strongest sign that it is not. Do not accept confidence, fluent wording, a high score, or a successful first attempt as proof. Write the additional check that would change the decision and name who owns that check.
How to use the prompts without outsourcing judgment
Before running any prompt, replace every placeholder, remove private information that is not required, and state which supplied sources the assistant may use. Save the initial input, output, corrections, and final human decision. This creates a record of how the tool contributed and makes it easier to spot when a later answer contradicts an earlier assumption.
- Benchmark designer: define the expected output before sending it, verify every material claim afterwards, and use this practical boundary: Run the same set across candidate routes.
- Cost calculator: define the expected output before sending it, verify every material claim afterwards, and use this practical boundary: Paste current official pricing.
- Migration test: define the expected output before sending it, verify every material claim afterwards, and use this practical boundary: For existing API users.
- Procurement summary: define the expected output before sending it, verify every material claim afterwards, and use this practical boundary: Useful before plan approval.
End by writing a short decision note in your own words: what you learned, what remains uncertain, which source or test carries the most weight, what you decided, and when the decision should be reviewed. That note is often more valuable than the original AI output because it captures accountable judgment rather than a temporary answer.
My 2026 recommendation
If a person wants to create and edit images conversationally, compare current ChatGPT plans based on the entire product and actual limits shown in the account. If software needs image generation, start with OpenAI's currently supported image models and calculate from the live documentation.
Keep DALL-E 3 in the article because people still search the name and existing integrations still exist. But label it honestly as legacy API technology, not the centre of a new buying guide.
Primary sources checked
Build the app you'll plug this AI into
Whichever plan you pick, you still need the product it powers. Lovable builds full-stack web apps from plain-English prompts, connect any AI tool or API and ship the same week. Free to start.
Affiliate link, we may earn a commission at no extra cost to you.
Plans & Pricing
ChatGPT Images
People creating and editing images interactively
LIMITATIONS
DALL-E GPT
Existing users who specifically need DALL-E
LIMITATIONS
DALL-E 3 API
Maintaining an existing integration during migration
LIMITATIONS
Current Image API
New applications and production workflows
LIMITATIONS
Hidden Costs to Watch For
A finished asset may require several generations and edits
Storage, moderation, retries and human review sit outside model pricing
ChatGPT plan limits and API rates can change
Existing DALL-E 3 API integrations need a migration plan
Our Verdict
For interactive creation, compare current ChatGPT plans and use ChatGPT Images. For a new software integration, use a currently supported OpenAI image model and calculate cost from the live documentation. Treat DALL-E 3 API access as a legacy migration case.
Alternatives to Consider
Frequently Asked Questions
Is DALL-E 3 API access still available?▼
Yes, but OpenAI marks it as deprecated and says it is slated for removal in the future. New integrations should use a currently supported image model.
Is ChatGPT Images the same as DALL-E 3?▼
No. OpenAI describes a current ChatGPT Images experience while continuing to offer DALL-E through a separate DALL-E GPT. Product access and limits depend on the current ChatGPT tier.
How should I estimate image API cost?▼
Use the live model pricing for the chosen quality and dimensions, then multiply by realistic attempts and edits per accepted asset. Add review, moderation, storage and retry costs.
Other AI Tool Pricing
Midjourney Pricing
Image Generation
🎯Stable Diffusion Pricing
Image Generation
🔥Adobe Firefly Pricing
Image Generation
🎨Leonardo AI Pricing
Image Generation