DALL-E 3 generated over 916 million images and held 24.35% of the AI image generation market before OpenAI officially retired it on May 12, 2026. The model, which powered image creation inside ChatGPT and Microsoft Copilot for over two years, has been replaced by OpenAI’s GPT Image model family. This article covers DALL-E 3’s usage data, market position, pricing history, technical specs, and the transition to its successor.
DALL-E 3 Statistics 2026 — TL;DR
DALL-E 3 accumulated 916 million generated images across all access channels through 2024, processing roughly 4 million images per day at peak usage.
The model held 24.35% of the AI image generation market, second only to Midjourney’s 26.8% share.
OpenAI deprecated both DALL-E 2 and DALL-E 3 APIs on May 12, 2026, directing developers to GPT Image 2, GPT Image 1, or GPT Image 1 Mini.
ChatGPT users were automatically migrated from DALL-E 3 to GPT Image 1.5 in December 2025, months ahead of the API shutdown.
API pricing ranged from $0.040 per standard 1024×1024 image to $0.120 per HD 1792×1024 image throughout its lifespan.
How Many Images Did DALL-E 3 Generate?
DALL-E 3 produced an estimated 916 million images from its October 2023 launch through the end of 2024. At peak output, the platform processed approximately 4 million images per day across ChatGPT Plus, the OpenAI API, and Microsoft Copilot. The earlier DALL-E 2 model had already hit 2 million daily generations before the upgrade.
For context, all AI image generators combined produce around 34 million images per day. The cumulative total across all platforms exceeded 15 billion images by late 2024. Stable Diffusion-based tools account for about 12.59 billion of that total due to widespread open-source deployment.
| Platform | Cumulative Images | Daily Output (Peak) |
|---|---|---|
| Stable Diffusion (all) | 12.59 billion | ~24 million |
| Midjourney | 964 million | ~3 million |
| DALL-E 3 | 916 million | ~4 million |
| Adobe Firefly | 1 billion+ | N/A |
Source: Everypixel Journal, Quantumrun Foresight
DALL-E 3 Market Share Compared to Competitors
Midjourney led the proprietary AI image generation market with 26.8% share as of 2024, followed by DALL-E at 24.35%. NightCafe held 23.2%, while Stable Diffusion captured 15.1% by user preference despite generating far more total images through its open-source ecosystem. The remaining 10.5% was split among smaller platforms like GetImg AI, Starry AI, and others.
Midjourney reported $500 million in revenue for 2025 — a 66.7% increase over 2024 — with a team of around 40 people and no outside funding. DALL-E 3 revenue figures were never disclosed separately from OpenAI’s broader financials.
DALL-E 3 User Base and Access Points
OpenAI reported 1.5 million active DALL-E users before integrating DALL-E 3 into ChatGPT in October 2023. That integration opened the model to tens of millions of ChatGPT Plus and Enterprise subscribers. By August 2025, ChatGPT itself had 700 million weekly active users, though only a fraction of those used image generation features regularly.
DALL-E 3 was accessible through three channels: ChatGPT (Plus at $20/month, Enterprise at custom pricing), the OpenAI API (pay-per-image), and Microsoft Copilot (formerly Bing Image Creator, free with limits). Free-tier ChatGPT users gained access in August 2024 with a cap of 2 images per day. Enterprise adoption reached roughly 70,000 businesses using DALL-E across their operations.
| Access Channel | Cost | Image Limits |
|---|---|---|
| ChatGPT Plus | $20/month | ~50 images per 3 hours |
| ChatGPT Free | $0 | 2 images per day |
| OpenAI API (Standard 1024×1024) | $0.040/image | Rate-limited |
| OpenAI API (HD 1792×1024) | $0.120/image | Rate-limited |
| Microsoft Copilot | Free | Daily cap applies |
Source: OpenAI API documentation, Microsoft
DALL-E 3 API Pricing History
DALL-E 3’s API pricing stayed fixed from its November 2023 API launch until deprecation. Standard quality at 1024×1024 cost $0.040 per image. HD quality at the same resolution was $0.080. The widest format — HD at 1792×1024 — ran $0.120 per image. No volume discounts were publicly available.
By comparison, GPT Image 2 (the current replacement) starts at $0.005 per image on its lowest quality tier. GPT Image 1 Mini matches that floor. At scale — say 50,000 images per month — DALL-E 3 standard would have cost $2,000. Self-hosted Stable Diffusion handles the same volume for around $200 in compute costs.
DALL-E 3 Technical Specifications
DALL-E 3 supported three output resolutions: 1024×1024 (square), 1792×1024 (landscape), and 1024×1792 (portrait). Both standard and HD quality tiers were available at each resolution. OpenAI never disclosed the exact parameter count for DALL-E 3, though its predecessors used 12 billion (DALL-E 1, based on GPT-3) and 3.5 billion (DALL-E 2, optimized diffusion).
The model’s main advantage over DALL-E 2 was prompt fidelity. DALL-E 3 used GPT-4 to automatically rewrite and expand user prompts before generating images, which reduced the need for manual prompt engineering. Text rendering inside images — a long-standing weakness of diffusion models — improved to roughly 4 out of 5 accuracy in independent tests. Midjourney scored 0 out of 5 on the same benchmarks for in-image text.
| Specification | DALL-E 2 | DALL-E 3 |
|---|---|---|
| Max Resolution | 1024×1024 | 1792×1024 |
| Parameters | 3.5 billion | Undisclosed |
| Quality Tiers | Standard only | Standard + HD |
| Text Rendering | Poor | Good (4/5 accuracy) |
| ChatGPT Integration | No | Yes (Oct 2023) |
| C2PA Watermarking | No | Yes (Feb 2024) |
Source: OpenAI, OnyxRanked (independent benchmarks, May 2026)
DALL-E 3 Deprecation Timeline
OpenAI began moving away from DALL-E 3 in March 2025, when GPT Image 1 (internally gpt-image-1) launched as a native image generation capability inside ChatGPT. Unlike DALL-E 3 — which functioned as a separate tool called by ChatGPT — GPT Image was built directly into the model’s output pipeline. This allowed it to maintain conversational context across multiple image iterations, something DALL-E 3 could not do.
In December 2025, ChatGPT users were automatically migrated to GPT Image 1.5. GPT Image 2 followed in April 2026 with a fully rebuilt architecture. On May 12, 2026, OpenAI pulled the plug on both DALL-E 2 and DALL-E 3 APIs. Any API calls using the dall-e-2 or dall-e-3 model strings now return errors. The recommended replacements are gpt-image-2, gpt-image-1, and gpt-image-1-mini.
| Date | Event |
|---|---|
| September 2023 | DALL-E 3 announced |
| October 2023 | DALL-E 3 integrated into ChatGPT Plus |
| November 2023 | DALL-E 3 API launched |
| February 2024 | C2PA watermarking added |
| August 2024 | Free ChatGPT users gain access (2 images/day) |
| March 2025 | GPT Image 1 replaces DALL-E 3 in ChatGPT |
| December 2025 | ChatGPT migrated to GPT Image 1.5 |
| April 2026 | GPT Image 2 released |
| May 12, 2026 | DALL-E 2 and DALL-E 3 APIs shut down |
Source: OpenAI Changelog, Wikipedia
AI Image Generation Market Size in 2026
The global AI image generator market grew from $11.65 billion in 2025 to an estimated $15.18 billion in 2026, a 30.3% year-over-year increase according to Research and Markets. A separate estimate from MarketsandMarkets puts the AI image generation segment at $12.4 billion for 2026, representing roughly 18% of the total generative AI market. Projections for 2030 range from $60.8 billion to $88.7 billion depending on the source and market definition.
North America accounts for about 41% of global AI image generation revenue. Over 150 million people use AI image generators monthly as of 2026, producing approximately 80 million images per day. Among surveyed professionals, 62% of marketers use generative AI for image creation. Enterprise adoption is concentrated in retail, e-commerce, media, and marketing sectors.
How Does DALL-E 3 Compare to GPT Image 2?
GPT Image 2, released April 21, 2026, is architecturally different from DALL-E 3. It no longer runs on the GPT-4o image pipeline. Where DALL-E 3 operated as an external API call triggered by ChatGPT, GPT Image 2 generates images natively within the model — the same way it produces text. This means it can reference earlier parts of a conversation, iterate on uploaded images, and maintain visual consistency across a session.
Pricing dropped substantially. GPT Image 2 starts at $0.005 per image at low quality (1024×1024), compared to DALL-E 3’s $0.040 floor. High quality at the widest resolution runs $0.211, compared to $0.120 for DALL-E 3 HD. The Studio Ghibli-style image trend in March 2025 — which went massively viral — was powered by GPT Image 1, not DALL-E 3, and drove ChatGPT past 150 million total users in that period.
| Feature | DALL-E 3 | GPT Image 2 |
|---|---|---|
| Lowest API Price | $0.040/image | $0.005/image |
| Max Resolution | 1792×1024 | 1536×1024 |
| Quality Tiers | Standard, HD | Low, Medium, High |
| Native ChatGPT | No (external call) | Yes (built-in) |
| Image Editing | No | Yes |
| Context Awareness | None across turns | Full conversation context |
Source: OpenAI API documentation, CostGoat (May 2026)
DALL-E 3 Content Authenticity and Safety
OpenAI added C2PA metadata watermarking to all DALL-E 3 outputs in February 2024. The provenance classifier achieved 98% accuracy on unmodified DALL-E 3 images and maintained above 95% accuracy after common modifications like cropping, resizing, or JPEG compression. Detection accuracy for images from competing generators like Midjourney reached only 5–10%, meaning the system was purpose-built for OpenAI’s own output.
Human ability to distinguish AI-generated images from real ones has dropped to around 38% accuracy as of 2026, according to Imagera AI research. This makes automated provenance tools like C2PA increasingly relevant for platforms dealing with AI-generated content at scale.
DALL-E 3 Statistics by Industry Use
Marketing and advertising accounted for over 36% of AI image generation spending in 2025. Media and entertainment followed at 34% of total market activity. E-commerce and retail adoption grew at 39% year-over-year, with AI tools saving e-commerce teams an average of 6.4 hours per week on visual content production.
Among individual creators, 29% of digital artists had integrated AI image tools into their professional workflows by the end of 2025. A survey of professional artists found 28.3% preferred DALL-E 2 as their tool of choice, compared to 27.1% for Midjourney — a surprisingly narrow gap given Midjourney’s larger overall market share. Among photo editing workflows, AI-powered tools saw 441% year-over-year growth in listings and traffic during 2024.
FAQ
Is DALL-E 3 still available in 2026?
No. OpenAI shut down both DALL-E 2 and DALL-E 3 APIs on May 12, 2026. ChatGPT users were migrated to GPT Image 1.5 in December 2025. GPT Image 2 is the current replacement.
How many images did DALL-E 3 generate in total?
DALL-E 3 generated approximately 916 million images through the end of 2024, with daily output reaching about 4 million images at peak usage across all access channels.
What was DALL-E 3’s market share?
DALL-E 3 held 24.35% of the AI image generation market as of 2024, ranking second behind Midjourney at 26.8%. NightCafe followed at 23.2%.
How much did DALL-E 3 cost per image?
API pricing ranged from $0.040 per standard 1024×1024 image to $0.120 for HD images at 1792×1024. ChatGPT Plus subscribers paid $20/month for roughly 50 generations every 3 hours.
What replaced DALL-E 3?
OpenAI’s GPT Image model family replaced DALL-E 3. GPT Image 1 launched in March 2025, GPT Image 1.5 in December 2025, and GPT Image 2 in April 2026. Pricing starts at $0.005 per image.
Sources
https://platform.openai.com/docs/changelog
https://journal.everypixel.com/ai-image-statistics
