OpenAI Rolls Out ChatGPT Images 2.5 With Faster Generation and Two New API Models
OpenAI has released Images 2.5, its latest image-generation system. The company says the update improves quality over the previous model. ChatGPT users already create more than three billion images each week. OpenAI now promises sharper detail, tighter edits, and quicker results for that workload.
The firm outlined the changes in a blog post on Tuesday. Lighting looks more natural, it says. Textures look richer. Latency falls by about 50 percent versus Images 2.0. New tools sit alongside that speed gain. Sketch, templates, and comments are the main additions.
Sketch lets people draw inside ChatGPT and use that drawing as a reference. Typing @sketch in the composer opens the tool. The aim is to turn a rough idea into a first concept without leaving the chat. Templates live in the Images tab. The system can ask follow-up questions after a user starts. That loop is meant to produce more tailored flyers, product shots, and other common formats.
Sharing works differently too. Users can publish a prompt with the finished image. Friends can then reuse the same text and generate their own version. OpenAI presents this as a simple way to pass a look around a team or a group.
ChatGPT Images 2.5 is rolling out to all ChatGPT, ChatGPT Work, and Codex accounts. Desktop, mobile, and web clients are included.
Developers get two separate API models at the same time. GPT-Image-2.5 Flare is the default for most apps. OpenAI says it beats GPT-Image-2 on quality while cutting latency by half. Typical uses include social posts, product mockups, visual search, rapid prototypes, and high-volume generation.
GPT-Image-2.5 Sunburst targets premium work instead. It follows multi-step edit instructions more closely. Campaign art and polished product imagery are the intended jobs. Generation takes longer. Control across successive edits is the trade-off.
Both API models can produce transparent backgrounds. That helps with assets for slides, games, and web layouts. OpenAI also continues to attach C2PA metadata and an invisible watermark so generated pictures can be identified later.
No launch-day price change was highlighted for ChatGPT users. API billing still runs on image tokens. Exact per-image cost will depend on quality settings and resolution.
The release keeps image tools inside the same ChatGPT window rather than a separate app. For now, the claim to test is simple. Pictures should look cleaner. They should arrive faster. Edits should stick more closely to what the user asked.