OpenAI simply shipped Photographs 2.6 throughout ChatGPT, Codex and the API. For builders, the core launch is GPT-Picture-2.6, shipped as two API fashions: Flare and Sunburst. OpenAI group states that Photographs 2.6 generates photographs 50% quicker than earlier than. The replace additionally provides GIF output, SVG export, markup-guided edits and a brand new complexity parameter.
TL;DR
- Measurement: Closed weights, token-billed picture API.
- Runs on: OpenAI-hosted solely (ChatGPT, Codex, API).
- Efficiency: 50% quicker picture era than the earlier launch, per OpenAI.
- Finest: Flare can return a picture in as little as 6 seconds at low high quality.
- Worst: No benchmark scores and no per-image value desk printed for two.6.
What’s GPT-Picture-2.6?
GPT-Picture-2.6 is OpenAI’s picture era and enhancing mannequin household for the API. It accepts textual content and picture inputs and returns photographs. GPT-Picture-2.6 Flare is constructed for quick iteration and interactive workflows. GPT-Picture-2.6 Sunburst is constructed for detailed compositions and enhancing.
OpenAI states that Flare can generate photographs in as little as 6 seconds when high quality is about to low. Sunburst targets greater high quality and stronger consistency throughout edits at medium or prime quality.
What modified in ChatGPT Photographs 2.6?
ChatGPT customers can now flip a immediate or a photograph into an animated GIF. They’ll alter animation pace, edit particular person frames and share the outcome. Customers may also save and tag reference photographs of individuals, pets and characters. These references could be reused later with out importing them once more.
Utilization retains climbing. OpenAI says folks now generate greater than 4B photographs per week throughout ChatGPT and the API. That’s up from greater than 3B per week 8 weeks earlier, on the Photographs 2.5 launch. On common, that’s over 500M photographs per day.
How does the brand new complexity parameter work?
The complexity area units how a lot the mannequin plans earlier than it renders. The picture era information lists 3 ranges: low, medium and excessive. When you omit it, each fashions use excessive.
Low fits centered edits, resembling altering a background or an indication. Excessive helps with dense textual content, complicated layouts and relationships between objects. OpenAI warns that greater complexity can add latency and introduce undesirable modifications. Adjusting complexity has no separate price. It’s separate from high quality, which controls rendering element and constancy.
window.addEventListener(“message”,perform(e){if(e.information&&e.information.id===”mtp-gi26″&&e.information.mtpH){doc.getElementById(“mtp-gi26-frame”).model.peak=e.information.mtpH+”px”;}});
Markup edits, SVG and GIF output within the API
Picture markup allows you to mark up the precise area you need modified. The launch submit says markup is coming to the API. The information already lists markup-guided edits as supported for each fashions. It additionally lists SVG output and animated GIF output via the Picture API.
SVG issues for icons, diagrams and UI property that should scale with out blur. GIF output removes a separate animation step for social and product groups.
How a lot does GPT-Picture-2.6 value?
Pricing is token-based and similar for Flare and Sunburst, per the mannequin pages:
- Textual content enter: $5.00 per 1M tokens ($1.25 cached)
- Picture enter: $8.00 per 1M tokens ($2.00 cached)
- Picture output: $30.00 per 1M tokens
How does GPT-Picture-2.6 evaluate with Nano Banana 2.1 and FLUX 3 Picture?
| Function | GPT-Picture-2.6 Flare / Sunburst | GPT-Picture-2.5 Flare / Sunburst | Google Nano Banana 2.1 | BFL FLUX 3 Picture |
|---|---|---|---|---|
| Launched | Nov 4, 2026 | Sep 8, 2026 | Oct 2026 | Oct 1, 2026 |
| Parameters | Not disclosed | Not disclosed | Not disclosed | Not disclosed (Klein: 5.6B) |
| Weights | Closed | Closed | Closed | Closed (Klein open, FLUX Kommunity License) |
| Planning management | complexity: low / medium / excessive | Not disclosed | Pondering: minimal / medium / excessive | Bounding containers in immediate |
| Max reference photographs | Not disclosed | Not disclosed | 14 | 10 |
| Max output | Not disclosed | Not disclosed | 4K | 4K (about 16 MP) |
| GIF output | Sure | Not disclosed | Not disclosed | Not disclosed |
| SVG output | Sure | Not disclosed | Not disclosed | Not disclosed |
| Pricing | $8 / $30 per 1M picture in / out | $8 / $30 per 1M picture in / out | $0.0336 per 1K picture | $0.048 per 1K picture |
| Pace declare | Flare: 6 s minimal at low high quality | Flare: as much as 50% decrease latency vs GPT-Picture-2 | Not disclosed | Not disclosed |
Google and BFL publish per-image costs. Nano Banana 2.1 begins at $0.0336 per 1K picture and reaches $0.113 at 4K. FLUX 3 Picture begins at $0.048 per 1K picture and reaches $0.607 at 4K. Its open-weight sibling, FLUX 3 Picture Klein, runs on client GPUs right down to an RTX 3090. OpenAI’s edge is native GIF and SVG output. Its hole is transparency on per-image value and high quality.
Key Takeaways
- Flare returns photographs in as little as 6 seconds at low high quality.
- Sunburst targets element and edit consistency at medium or prime quality.
- New
complexityparameter (low, medium, excessive; default excessive) prices nothing additional. - Token costs match GPT-Picture-2.5: $30 per 1M picture output tokens.
Take a look at the launch submit, the Flare mannequin card, the Sunburst mannequin card and the picture era information. Additionally, be happy to comply with us on Twitter and don’t neglect to hitch our 150k+ML SubReddit and Subscribe to our Publication. Wait! are you on telegram? now you’ll be able to be a part of us on telegram as effectively.
Must associate with us for selling your GitHub Repo OR Hugging Face Web page OR Product Launch OR Webinar and many others.? Join with us
The submit OpenAI Launches GPT-Picture-2.6 Flare and Sunburst: Provides GIFs, SVG Export and 6-Second API Technology appeared first on MarkTechPost.
