OpenAI Releases Images 2.6 With GIF Support and Faster Generation
OpenAI has released Images 2.6 globally across ChatGPT, Codex, and its API, delivering a 50% speed increase over the previous version. The update introduces two API models, Flare for rapid generation in as little as six seconds and Sunburst for higher fidelity, plus native GIF animation from sprite-sheet prompts, image markup and SVG exports in the API, and a complexity parameter to balance rendering effort against quality.
Short answer: OpenAI has released Images 2.6 globally across ChatGPT, Codex, and its API, delivering a 50% speed increase over the previous version. The update introduces two API models, Flare for rapid generation in as little as six seconds and Sunburst for higher fidelity, plus native GIF animation from sprite-sheet prompts, image markup and SVG exports in the API, and a complexity parameter to balance rendering effort against quality.
OpenAI Images 2.6 release ChatGPT Codex API
OpenAI just dropped Images 2.6 across ChatGPT, Codex, and the API, the newest bump for their image stack. It's live globally today for every tier on desktop, mobile, and web, packing four big changes: faster generations, native GIFs, tighter editing controls, and fresh API knobs so developers can actually balance cost against quality.
The scale here is wild. OpenAI says users crank out roughly four billion images a week across their platforms, that's over half a billion a day. With that kind of volume, the speed gains matter. They're reporting a fifty percent jump over the last version. In the API, you've got two flavors now: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. Flare is the speed demon; it can spit out an image in as little as six seconds on low quality, and it's noticeably better at layouts for posters, product shots, that sort of professional work. Sunburst leans the other way, higher fidelity, stronger consistency when you're iterating at medium or high quality.
GIF support motion images OpenAI 2.6
If you need motion, the GIF feature is a nice touch. You can animate a brand mascot, whip up reaction memes, or storyboard character movement before you commit to full video. Feed it a single prompt describing a sprite sheet, basically a grid of sequential frames, and it hands back a seamless loop. Early demos show a chibi office worker freaking out over a crashing stock chart and another celebrating a rally, both from detailed prompts nailing frame count, grid layout, and style. Even a rough hand-drawn sketch turns into an expressive animated character, which drops the floor for creators without polished assets.
Editing gets sharper, too. Image markup, already in ChatGPT, lands in the API so devs can pinpoint exactly which region needs changing. SVG exports mean the output is web-, app-, and print-ready without extra conversion steps. A new complexity parameter controls how much "thinking" the model does before rendering: low for quick tweaks, medium or high when a scene is packed with elements or constraints. OpenAI suggests pairing low complexity with whatever quality level your workflow needs for most edits, then dialing it up only when the scene actually demands it.
Images 2.6 pricing system card safety evals
Pricing for both models is up on the dev site, and there's a system card detailing safety evals and safeguards for 2.6. For builders, the takeaway is a clearer menu of speed-versus-quality options, native animation output, and programmatic editing tools that slot into existing pipelines. For everyday folks in ChatGPT and Codex, it should just feel snappier and more versatile, especially when you're iterating on visual concepts or adding motion to static ideas.
Optimize image generation latency quality knobs
Anyone baking image gen into a product should poke at the new complexity and quality knobs to see where they can shave latency without tanking the fidelity their use case needs. Teams waiting on API-side markup or SVG output can start migrating those workflows now. And creators itching to play with animated assets have a low-friction path from sketch to looping GIF without leaving the platform.
Frequently asked questions
What are the two new image generation models in OpenAI's Images 2.6 API?
The two models are GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. Flare prioritizes speed, generating images in as little as six seconds on low quality and excelling at professional layouts like posters and product shots. Sunburst prioritizes higher fidelity and stronger consistency when iterating at medium or high quality settings.
How does the new GIF animation feature work in Images 2.6?
Users provide a single prompt describing a sprite sheet, a grid of sequential frames, and the model returns a seamless looping GIF. This enables animating brand mascots, creating reaction memes, or storyboarding character movement. Early demos show detailed prompts specifying frame count, grid layout, and style, and even rough hand-drawn sketches can become expressive animated characters.
What editing improvements does Images 2.6 bring to the API?
Image markup, previously only in ChatGPT, is now available in the API so developers can pinpoint exact regions to change. SVG exports make output web-, app-, and print-ready without extra conversion. A new complexity parameter controls how much "thinking" the model does before rendering: low for quick tweaks, medium or high for scenes packed with elements or constraints.
How much faster is Images 2.6 compared to the previous version?
OpenAI reports a fifty percent speed jump over the last version. With users generating roughly four billion images per week, over half a billion daily, this latency reduction is significant. In the API, the Flare model can produce an image in as little as six seconds on low quality, while Sunburst offers higher fidelity at medium and high quality tiers.
Where can developers find pricing and safety information for Images 2.6?
Pricing for both GPT-Image-2.6 Flare and Sunburst models is published on OpenAI's developer site. A system card detailing safety evaluations and safeguards for version 2.6 is also available there. Developers can use the new complexity and quality knobs to balance latency against fidelity for their specific use cases.
Source: OpenAI
Related posts
Texas Blue Wave Imperils Massive AI Data Center Projects
Democrats swept Texas' Senate, governorship, and House in November midterms, threatening massive AI data center projects. Governor-elect Gina Hinojosa campaigned for a total moratorium, while outgoing Gov. Abbott froze permits in September. Projects like the 6,400-megawatt Nexus Hubbard and 1,820-megawatt Stargate Frontier face delays. Banks are already tightening lending, though local officials and construction unions may resist statewide restrictions.
US Voters Back AI Restrictions in 2026 Midterms
Voters rebuked AI expansion in the 2026 midterms. Across ten states, 38 of 44 data center restriction measures passed. Democrat Gina Hinojosa won Texas's governorship, jeopardizing 13 gigawatts of planned capacity. Democrats also won Arizona, Georgia, and Ohio, threatening two-thirds of 2028's capacity. NBC exit polls showed three in four voters opposed data centers, with over half citing AI as crucial to their congressional vote. Analysts warn permitting will tighten, raising costs an
OpenAI Pipeline Finds 948 Linux Kernel Privilege Escalation Flaws
OpenAI built an autonomous pipeline discovering 948 working local privilege escalation exploits in the Linux kernel. The system identified 2,308 candidates, reproduced 1,724, and generated exploits validated by humans before reporting to maintainers with patches. The five-stage process used Scout agents to target 2.8 million code lines, analyst agents to find candidates, and triage agents to create proofs of concept. Exploit agents tested against standard kernel protections, confirming