Back to changelog

GPT Image 2.6 With Native SVG Output

OpenAI's GPT Image 2.6 Sunburst and Flare are live, and /v1/images/generations now returns vector graphics with output_format: svg. Generate icons, logos, stickers, and illustrations as editable SVG through your existing gateway key.

LLM Gateway card announcing GPT Image 2.6 with native SVG output

Icons, logos, and stickers need to scale, recolor, and ship as code. A PNG from an image model means a second tool to trace it into vectors. GPT Image 2.6 draws the vectors itself: set output_format: "svg" and the gateway returns an SVG document you can drop into a page or open in a design tool.

Two new models

Model Use it for
openai/gpt-image-2.6-sunburst Detailed compositions and precise edits
openai/gpt-image-2.6-flare Fast iteration on generation and editing

Both accept text and image inputs, support low, medium, high, xhigh, max, and auto quality, and work on /v1/images/generations, /v1/images/edits, and /v1/chat/completions.

Generate an SVG

1curl https://api.llmgateway.io/v1/images/generations \2  -H "Authorization: Bearer $LLM_GATEWAY_API_KEY" \3  -H "Content-Type: application/json" \4  -d '{5    "model": "openai/gpt-image-2.6-flare",6    "prompt": "A flat two-color sticker of a smiling cactus",7    "output_format": "svg",8    "svg": { "model": "gpt-5.4-mini", "reasoning": { "effort": "low" } }9  }' | jq -r '.data[0].b64_json' | base64 --decode > cactus.svg

data[0].b64_json holds the base64-encoded SVG, and the response echoes "output_format": "svg". The document is static vector content: no scripts, external resources, or embedded raster images.

  • When your OpenAI account has SVG model selection enabled, supply svg.model as above; otherwise omit svg. Author reasoning effort is optional.
  • n must be 1; output_compression and streaming are not available for SVG.
  • SVG is a generation format. /v1/images/edits returns raster images.
  • A model without SVG output gets a 400 before any provider is called.
  • DevPass coding plans cover text models only, so image requests from them return a 403.

/v1/images/generations also accepts png, jpeg, and webp through output_format, plus output_compression for the last two.

Pricing

GPT Image 2.6 uses the same token rates as GPT Image 2 and 2.5. For SVG output, OpenAI first renders a reference image, then an authoring model writes the vectors. Both are billed from the reported usage: image tokens at the image model's rates and the authoring tokens at the authoring model's catalogue rates, shown in usage.cost. Unpriced authoring cost is omitted and logged for reconciliation; the reference image is still billed. See the models page for current prices.

SVG output docs → | Generating SVGs with AI →