ChatGPT Image Generation: Complete Model Family Guide
From DALL-E to ChatGPT Images 2.5: the complete guide to OpenAI's image generation model family, covering architecture, versions, and how to choose.
Before reading, create with ChatGPT Images 2.5.
Developer: OpenAI · Family launched: January 2021 · Current-generation model: ChatGPT Images 2.5 · Architecture: Not fully specified publicly · Official documentation: ChatGPT Images 2.5 announcement
What Is the ChatGPT Image Model Family?
OpenAI has shipped multiple generations of image generation technology since January 2021. The progression spans three product eras: the original DALL-E research model, diffusion-based DALL-E 2 and DALL-E 3, and the later GPT Image family that powers current ChatGPT image workflows. OpenAI has not published a complete architecture specification for every deployed GPT Image version.
The transition from DALL-E 3 to GPT Image 1 in March 2025 was a new product generation rather than a minor DALL-E revision. OpenAI positioned GPT Image around stronger multimodal prompting, editing, and text handling; the public documentation does not establish a single architecture explanation or universal quality score for the whole family. OpenAI retired DALL-E 2 and DALL-E 3 on May 12, 2026.
The cultural signal came immediately. When GPT Image 1 launched inside ChatGPT in March 2025, Studio Ghibli-style portrait transformations became a widely shared example of image-to-image editing. With GPT Image 2 in April 2026, the product added Reasoning Mode for pre-generation planning and output verification.
The DALL-E Era: 2021–2023
OpenAI's image generation technology began in January 2021 with the original DALL-E, a 12-billion-parameter variant of GPT-3 that compressed images into discrete tokens using a discrete variational autoencoder (dVAE) and predicted them sequentially alongside text. DALL-E demonstrated for the first time that language models could generate coherent images from text descriptions, such as "an armchair in the shape of an avocado."
In April 2022, OpenAI shifted to diffusion architecture with DALL-E 2 (unCLIP), combining CLIP embeddings with diffusion decoders to achieve 1024×1024 output resolution and commercially available inpainting. DALL-E 3 (October 2023) further refined this with synthetic recaptioning, training the model on detailed AI-generated image descriptions to dramatically improve prompt adherence. Despite these advances, DALL-E 3 shared the structural limitation of all diffusion architectures: text inside images remained statistically approximate rather than linguistically constructed, leaving spelling garbled across complex compositions. Both DALL-E 2 and DALL-E 3 were retired on May 12, 2026.
The GPT Image Era: 2025–Present
GPT Image 1 & 1.5: The Autoregressive Pivot (2025)
On March 25, 2025, OpenAI introduced GPT Image 1 inside ChatGPT as a new generation of image models. The release moved the product beyond the earlier DALL-E generation and emphasized prompt following, multimodal inputs, and text in images. A Studio Ghibli-style portrait filter became a widely shared example of image-to-image editing, but individual viral examples are not controlled benchmarks.
Nine months later, in December 2025, OpenAI released GPT Image 1.5 with faster generation and more precise editing controls. Both versions laid the foundation for the current generation and are now in their deprecation window: GPT Image 1 is scheduled to shut down on October 23, 2026, while GPT Image 1.5 is scheduled for December 1, 2026. For technical specs, API model IDs, pricing history, and migration parameters, see the complete GPT Image 1 and 1.5 reference guide.
GPT Image 2: The Previous API Generation (2026)
On April 21, 2026, OpenAI released GPT Image 2 (gpt-image-2), the previous API generation before GPT Image 2.5. The current model page describes it as a state-of-the-art model for fast, high-quality image generation and editing, with text and image inputs, image outputs, flexible sizes, high-fidelity image inputs, inpainting, and Batch API support.
GPT Image 2 remains useful as a comparison and migration baseline, but it should not be treated as the current ChatGPT product name. For version-specific output parameters and migration context, see the ChatGPT Image 2 complete model guide.
ChatGPT Images 2.5: Reference Fidelity and Workflow Control (2026)
On September 8, 2026, OpenAI announced ChatGPT Images 2.5 as the latest product release in the family. It focuses on preserving reference subjects, making more precise edits, maintaining consistency across multiple turns, and reducing generation latency versus Images 2.0. The release also adds Sketch, Templates, image comments, and prompt sharing in ChatGPT. For the version-specific changes and the two new API models, see the ChatGPT Images 2.5 guide.
The Model Shift: From DALL-E to GPT Image
Understanding this transition is useful context for why the GPT Image family behaves differently from DALL-E, especially for multimodal prompts and edits.
How Diffusion Models Work (and Why They Fail at Text)
Diffusion models begin from a field of random noise and iteratively denoise toward a coherent image across many refinement steps. All regions of the image are processed simultaneously on each pass; the model learns to reverse a mathematically defined noise process, gradually shaping pixels into structure and detail. This approach works well for photographic subjects because natural textures (skin, fabric, foliage, stone) are statistically learnable patterns that diffusion handles gracefully.
Text rendering is where diffusion architectures fail structurally. Characters appearing inside a generated image represent a few percentage points of the total pixel budget. A diffusion model encounters them the same way it encounters any other visual element: by learning what they statistically look like, not what they mean. It has no representation of the correct character sequence that spells a word; it approximates the visual shape of letters. Spelling errors, inverted characters, and plausible-looking nonsense are not training deficiencies that more data or compute can eliminate; they follow directly from the architecture. DALL-E 3's synthetic recaptioning improved many areas but could not resolve this structural gap.
What Is Publicly Known About GPT Image
GPT Image 1 introduced a different generation approach from the earlier DALL-E products. OpenAI has not published a complete architecture specification for every GPT Image version, so implementation details should be treated as technical interpretation rather than a guaranteed description of the deployed model.
Claims about shared text and image representations or token-generation order should therefore be read as analysis, not as an official architecture guarantee. The practical takeaway is simpler: test text accuracy, instruction following, edit boundaries, and consistency on the prompts your workflow actually uses.
Practical Consequences for Output Quality
Text rendering: GPT Image models improve the handling of text and complex instructions compared with earlier generations, but exact spelling, placement, and legibility still need to be checked in every production asset. The current public model documentation does not establish a universal character-level accuracy score for the family.
Instruction following on complex prompts: Later GPT Image releases are positioned for stronger handling of complex visual instructions, but the exact result still depends on the prompt, references, dimensions, and quality setting. Test spatial relationships and required attributes on representative inputs.
Multi-turn editing coherence: GPT Image workflows support natural-language edits such as changing a background, removing a figure, or adjusting a headline. Repeated edits can still change details that should remain fixed, so preserve the original references and inspect each result before approval.
All Versions at a Glance
| Version | Released | Architecture | Max Resolution | Notable Features | Status |
|---|---|---|---|---|---|
| DALL-E | Jan 2021 | Autoregressive (GPT-3 + dVAE) | 256×256 | First text-to-image proof of concept | Retired |
| DALL-E 2 | Apr 2022 | Diffusion (unCLIP + CLIP) | 1024×1024 | Public API, inpainting, outpainting | Retired May 2026 |
| DALL-E 3 | Oct 2023 | Diffusion + GPT-4 recaptioning | 1024×1024 | Synthetic recaptioning, prompt adherence | Retired May 2026 |
| GPT Image 1 | Mar 25, 2025 | Not fully specified publicly | 1024×1024 class | Multimodal image generation and editing | Scheduled shutdown Oct 2026 |
| GPT Image 1.5 | Dec 16, 2025 | Not fully specified publicly | 1024×1024 class | Faster generation and precision editing | Scheduled shutdown Dec 2026 |
| GPT Image 2 | Apr 21, 2026 | Not fully specified publicly | Up to 4K custom size | Image generation and editing, inpainting, Batch API | Previous API generation |
| GPT Image 2.5 Flare | Sep 8, 2026 | Not fully specified publicly | Up to 4K custom size | Speed-optimized generation and editing, xhigh/max quality | Active |
| GPT Image 2.5 Sunburst | Sep 8, 2026 | Not fully specified publicly | Up to 4K custom size | Quality-optimized generation and editing, xhigh/max quality | Active |
Source: OpenAI deprecation schedule. “Up to 4K custom size” follows the current image-generation guide's documented bounds; resolutions above 2,560×1,440 are experimental.
Which Model Should I Use?
As of September 2026, DALL-E 2 and DALL-E 3 are retired. GPT Image 1 is scheduled to shut down on October 23, 2026, while GPT Image 1.5 and GPT Image 1 Mini are scheduled to shut down on December 1, 2026. GPT Image 2.5 is the current product/API generation described in the release and developer documentation.
For new workflows: Start by testing GPT Image 2.5 Flare when speed and throughput matter, or GPT Image 2.5 Sunburst when demanding generation and editing quality matters more. Keep the prompt, references, dimensions, and quality setting fixed during the first comparison.
For legacy API integrations: If existing code references gpt-image-1 or gpt-image-1.5, schedule and validate a migration to gpt-image-2 before the applicable shutdown date. The key change to plan for is the pricing model: GPT Image 1.x used flat per-image rates, while gpt-image-2 uses token-based billing (image input: $8/M tokens, cached image input: $2/M tokens, image output: $30/M tokens). The API endpoint structure is compatible. See the GPT Image 1 and 1.5 migration guide for a complete parameter-level comparison.
Platform Availability
ChatGPT: The September 2026 release announcement describes ChatGPT Images 2.5 as the current product release and lists availability across ChatGPT surfaces. Exact access and rollout can depend on the product surface and account.
OpenAI API: GPT Image 2.5 Flare and Sunburst use the Image API for image generation and editing. They can also be selected inside the Responses API image_generation tool. GPT Image 2 remains available as a previous API generation; its current model page lists Image API generation/editing and Batch support, but not top-level Responses or Chat Completions support. Check the current OpenAI image generation guide for request parameters, output limits, moderation, and cost details.
Third-party platforms: VioEvo provides ChatGPT Images 2.5 workflows for text-to-image generation and image editing without requiring a direct OpenAI account. Generations consume platform credits under VioEvo's billing rules.
Frequently Asked Questions
Can ChatGPT generate images directly from text?
Yes. ChatGPT Images 2.5 is the current product release described by OpenAI. You can describe a scene, request visual changes, or use reference images where the product surface supports them. API integrations use the separate model IDs documented for GPT Image 2.5 Flare and Sunburst.
What is the difference between DALL-E and the GPT Image model family?
DALL-E 2 and DALL-E 3 were earlier diffusion-era products. GPT Image is a later model family with stronger multimodal prompting and editing behavior. OpenAI has not published a complete architecture specification or universal text-accuracy benchmark for every GPT Image version, so compare models using the prompts, references, and output checks relevant to your workflow.
Are DALL-E 2 and DALL-E 3 still available in ChatGPT or via API?
No. OpenAI officially retired DALL-E 2 and DALL-E 3 on May 12, 2026. GPT Image 1 and GPT Image 1.5 remain available during their deprecation windows, with shutdown dates in October and December 2026 respectively.
Which image generation model powers ChatGPT today?
ChatGPT's current product release is ChatGPT Images 2.5. The API model IDs gpt-image-2.5-flare and gpt-image-2.5-sunburst are separate from the VioEvo selector chatgpt-image-2-5.
Can I use OpenAI's image generation model without a ChatGPT subscription?
Yes. You can access GPT Image models through the official OpenAI API, subject to the model's current availability and account requirements. VioEvo provides a separate credit-based interface for supported image generation and editing workflows.
VioEvo supports ChatGPT Images 2.5 for text-to-image generation and image editing. No OpenAI account required.
