What is GPT Image 2.6 Sunburst?
A polished campaign image has to get several things right at once: the product, the scene, the wording and the relationship between them. In this QVeris guide, we examine Sunburst for those detail-sensitive briefs, using official examples and a transparent cost comparison.
OpenAI introduced Images 2.6 on November 4, 2026, with two API models: Flare and Sunburst. Its announcement reports 50% faster image generation. Flare targets fast iteration; Sunburst emphasizes detail and edit consistency at medium or high quality. The speed claim is OpenAI’s release statement, not a measured QVeris response time. Read the announcement
Our starting point: choose Sunburst when preserving product identity and controlling a complex composition matter more than producing many rough options. Define acceptance criteria before the first request.
Generation, editing, SVG and GIF
The API ID is gpt-image-2.6-sunburst. It accepts text and images and produces images. OpenAI documents image generation and image editing, with SVG and animated GIF workflows described in the image guide. Model specification
Use quality and complexity for different decisions
Quality sets rendering detail; complexity controls planning. A simple background replacement does not need the same planning as a scene with several products, exact text and positional constraints. OpenAI recommends low complexity for most edits.
Choose the output for the deliverable
Use still images for product pages, SVG for scalable graphic assets, and GIF for short looping motion. GIF is documented for generation, not the edit endpoint. For reference edits, identify the target image and state which product features must remain unchanged. Formats and request guidance
Sunburst image comparison: one brief, three planning levels
The official Sunburst football illustration uses the same prompt and medium quality for all three images. The changing variable is complexity: low, medium and high. This makes it useful for examining how scene planning changes the result. OpenAI official example
Compare the brief, not just visual richness
Read the headline, follow the trophy and inspect the arrangement of people. For a real campaign, score those requirements before deciding whether more planning was worthwhile. These are selected official outputs; they contain no latency or billing comparison.
Flare vs Sunburst: the same SVG brief
Both images below come from OpenAI’s bakery-logo example. Compare the treatment of the same brand brief: lettering, decorative elements and balance at a small display size. This is a visual comparison of two selected outputs, not a benchmark ranking. OpenAI official example
For Sunburst, decide how much ornament the brand can sustain. A detailed mark still needs a simpler version for an avatar or small label; inspect the vector paths and text before handing it to a designer.
A Sunburst GIF example for short loops
For social content, a short motion can give a character or product another expression. OpenAI’s Sunburst lantern example uses 9 frames at 200 ms per frame: a 1.8-second loop. Expand it to inspect the movement and the transition back to the first frame. OpenAI official example
Play the official Sunburst GIF
Treat motion review separately from still-image approval. Check silhouette consistency, camera stability and the loop seam. More frames also change the workload; a GIF is not priced by assuming it costs the same as one still.
Flare vs Sunburst API cost comparison
Flare and Sunburst currently share the following OpenAI API rates, in USD per million tokens. The meaningful price comparison is therefore the actual workload and number of accepted images—not a difference in these list rates. Flare official pricing · Sunburst official pricing
| Billing item | Flare | Sunburst |
|---|---|---|
| Text input | $5.00 | $5.00 |
| Cached text input | $1.25 | $1.25 |
| Image input | $8.00 | $8.00 |
| Cached image input | $2.00 | $2.00 |
| Image output | $30.00 | $30.00 |
The same assumed usage, the same calculated cost
Assume 1,000 text-input tokens and 1,000 image-output tokens per request. For an edit, add 2,000 image-input tokens. The table below applies the listed rates to those assumptions; it is not an image quote or a recorded model bill.
| Assumed workload | Flare (USD) | Sunburst (USD) |
|---|---|---|
| 1 generation | $0.035 | $0.035 |
| 1 edit, uncached input | $0.051 | $0.051 |
| 100 edits, uncached input | $5.10 | $5.10 |
| 100 edits, all input billed at cached rates | $3.525 | $3.525 |
Uncached edit: (1,000 × $5 + 2,000 × $8 + 1,000 × $30) / 1,000,000 = $0.051.
Cached pricing only applies when the service reports eligible cached input. Output remains charged at the output rate. If 100 edit requests produce 50 accepted images, the uncached example becomes $5.10 / 50 = $0.102 per accepted image. Actual dimensions, quality and token usage can change the result.
QVeris settles model calls in Credits. Neither 2.6 model currently has a verified public QVeris price in our catalog snapshot. We therefore keep the USD comparison above separate from QVeris pricing. Check current models and prices
Build a repeatable creative workflow
Three useful starting briefs
Brand key visuals: define hierarchy and exact copy. Reference-based product edits: preserve the product while changing the setting. Social animation: approve character identity before checking motion. These are different deliverables and should have separate acceptance criteria.
A QVeris brief you can adapt
Using the supplied product photograph, create a premium skincare campaign. Keep the bottle silhouette, label text and cap unchanged. Place it to the right on a pale stone surface, with three leaves behind it and empty space on the left for copy. Maintain one consistent light direction. Do not add claims or new labels.
Explore models and start with QVeris
At QVeris, we help you move from reading about a model to choosing a usable model and connection method. Explore our model catalog for current capabilities and pricing, use Playground to try available models, and consult the Model Gateway guide when your application is ready for API integration.
Bring your next AI workflow to QVeris
Explore the platform, or open the official experience page and choose an available model.
Questions before you choose
Is ChatGPT Images 2.6 an API model ID?
ChatGPT Images 2.6 is the product name. API requests use the exact Flare or Sunburst ID. A ChatGPT subscription and API usage are different purchases.
Is Flare cheaper than Sunburst?
Their published token rates match. Compare actual tokens, retries and accepted assets for the same brief before drawing a cost conclusion.
Are the images QVeris test results?
The labeled comparison images are original examples from OpenAI documentation. Our workflow recommendations are editorial guidance; we do not present those assets as QVeris test output.
Can I use these two models on QVeris today?
They are not present in the public catalog checked for this guide. You can explore QVeris and try currently supported models; check the live catalog for subsequent availability and prices.
Sources and related reading
Model and pricing sources checked November 5, 2026. Images retain their documented model labels; the banner uses OpenAI’s Sunburst football examples at three complexity levels.
OpenAI · Introducing Images 2.6
OpenAI · Flare specifications and pricing
OpenAI · Sunburst specifications and pricing
OpenAI · Image generation guide and examples