November 6, 2026. OpenAI's new image models, GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst, cost the same per token as GPT-Image-2.5: $8 per million image input tokens and $30 per million image output tokens. OpenAI announced Images 2.6 on November 4 with animated GIFs, SVG export and a new complexity setting, and says it generates images 50% faster. The headline price did not move, so what changes for your bill is how many tokens each image uses, which OpenAI has not published.
Key numbers
| Item | Number |
|---|---|
| Image input tokens, 2.6 Flare and Sunburst (same as 2.5) | $8.00 per 1M |
| Cached image input (same as 2.5) | $2.00 per 1M |
| Image output tokens (same as 2.5) | $30.00 per 1M |
| Text input tokens (cached $1.25) | $5.00 per 1M |
| Generation speed vs before (per OpenAI) | 50% faster |
| Flare fastest low-quality generation (as little as) | 6 seconds |
| Images created per week (up from 3 billion in September per The Verge) | 4 billion |
| Default complexity if omitted (no separate fee for complexity) | High |
OpenAI launch post (November 4, 2026), OpenAI API pricing page and image generation guide, and The Verge (November 4, 2026), all read on November 6, 2026.
What OpenAI announced
- Two API models. OpenAI's launch post says Flare can generate "in as little as six seconds" at low quality, while Sunburst targets detail and consistency across edits at medium or high quality.
- GIFs. ChatGPT can create a GIF from a prompt or a photo, and The Verge reports you can change the animation speed and edit individual frames. OpenAI's API guide lists animated GIF output through the Image API.
- SVG export and markup. Image markup, already in ChatGPT, comes to the API so a team can pinpoint where a change should happen. SVG exports are meant for websites, apps and print.
- A complexity setting. It controls how much the model plans before rendering, with levels low, medium and high.
- Availability. Images 2.6 is available in ChatGPT and Codex on all subscription tiers, and OpenAI says it creates more than half a billion images a day.
What the 2.6 models cost
The OpenAI API pricing page lists both 2.6 models at the same rates as both 2.5 models, per million tokens:
| Token type | Input | Cached input | Output |
|---|---|---|---|
| Image tokens | $8.00 | $2.00 | $30.00 |
| Text tokens | $5.00 | $1.25 | Not listed |
Flare and Sunburst carry identical rates, so choosing between them changes speed and fidelity, not the price per token. The pricing page points readers to a calculator in the image generation guide for per-image estimates. We did not find a tokens-per-image figure for 2.6 on either page, so we cannot give you a cost per image.
The complexity setting is free, but the default is high
OpenAI's guide says the complexity levels for both models are low, medium and high, and that "if you omit complexity, both models use high". It also says higher complexity "can increase latency and introduce unwanted changes", and that adjusting complexity "has no separate fee". A team that never sets the field gets the highest planning effort. OpenAI recommends low complexity for faster edits. We suggest testing low complexity on routine edits such as swapping a background or a sign, and keeping high for dense layouts.
How fast usage is growing
OpenAI says people now create 4 billion images a week across ChatGPT, Codex and the API. The Verge notes that is up from 3 billion a week reported in September, a rise of about 33%. At 4 billion a week, that is roughly 571 million a day, which fits OpenAI's "more than half a billion" a day.
What it means for small teams
If you already run GPT-Image-2.5 in a workflow, the token rates do not change when you switch the model name to 2.6. Three things are worth checking before you do. First, run the same ten prompts on both models and compare the token usage the API returns, because that is the number that sets your cost. Second, set complexity explicitly instead of accepting the high default. Third, decide whether GIF output and SVG export replace a design or motion step you now pay for. Our earlier note on GPT-Image-2.5 cost per image explains why a token rate is not a per-image price.
Teams that connect image generation to ad, product or social workflows can have it set up as part of an AI automation agency build, with usage limits and review steps around it.
What we could not confirm
We did not check the Batch tab of the pricing page for the 2.6 models, and the guide's code examples still use GPT Image 2.5. Check OpenAI's pricing page before you budget, since these pages change. Sources read November 6, 2026.
Frequently Asked Questions
OpenAI's API pricing page lists both GPT-Image-2.6 Flare and Sunburst at $8 per million image input tokens, $2 per million cached, and $30 per million image output tokens, plus $5 per million text input tokens. That is the same as the 2.5 models. OpenAI has not published a tokens-per-image figure that we could find, so the price per image depends on what your own calls consume.
OpenAI says Flare is built for speed, generating in as little as six seconds at low quality, and improves layout for posters and product imagery. Sunburst is built for detail and for consistency across edits at medium or high quality. Both are priced the same per token.
OpenAI's image generation guide says both 2.6 models support animated GIF output through the Image API. In ChatGPT, The Verge reports you can adjust a GIF's speed and edit individual frames. We did not find separate GIF pricing in the pages we read, so check OpenAI's pricing page before budgeting GIF output.
No. OpenAI's guide says adjusting complexity has no separate fee. If you leave it unset, both models use high, which can raise latency, so set it to low for simple edits.
OpenAI says Images 2.6 is available in ChatGPT and Codex across all subscription tiers, on desktop, mobile and web. Usage limits per plan were not stated in the launch post.