How much does one image cost?

It is billed by tokens: $5 input, $1.25 cached input, and $30 output per 1M tokens, the same as 2.5. Output tokens grow with quality: at 1024², low uses about 820 (about $0.0246 per image) and high about 1,760 (about $0.0528); low / medium use more than 2.5, high and above the same. Actual charge = official price × group multiplier.

How do I get SVG or an animated GIF?

Set output_format to svg to get an SVG (n must be 1). Set it to gif for an animation: frame_count sets the number of frames (2–16, default 9) and frame_duration_ms the time per frame (20–2000 ms, default 200). For GIF, size is the size of one frame, so use auto or a small size such as 512x512: all frames are tiled into one sheet no larger than 3840 per side, and large frames fail.

What does complexity do?

It sets how much the model plans before rendering: low / medium / high, default high, with no separate fee. Any other value returns 400.

Can I edit by drawing annotations?

Yes. Send markup to /v1/images/edits: a transparent PNG the same size as the first input image with circles, arrows, or notes drawn on it (a multipart file part, or a URL or Data URL in JSON), and the model edits as annotated. markup cannot be combined with mask.

gpt-image-2.6-flareImageExamples verified 2026-09-25

GPT Image 2.6 Flare

The speed-focused GPT Image 2.6 model: about 20 seconds at 1024² on high, with new SVG and animated GIF output.

GPT Image 2.6 Flare (gpt-image-2.6-flare) is the speed-focused GPT Image 2.6 model, released by OpenAI on 2026-11-04: about 14 seconds at 1024² on low and about 20 seconds on high, with five quality tiers from low to max, plus the new complexity parameter, SVG output, and animated GIF output. It is billed by tokens at an official price of $5 input and $30 output per 1M tokens; at 1024², low uses about 820 output tokens (about $0.0246 per image) and high about 1,760 (about $0.0528). Call POST /v1/images/generations or /v1/images/edits.

Official price: $5 / $30/ 1M tokensInput / output
Size: Longest side ≤ 3840Sides in multiples of 16 · aspect ratio ≤ 3:1
Total pixels: 0.66–8.29 MP
Images per request: 1–10
Reference images: Up to 16Remote URLs ≤ 25 MiB each

Notes

size accepts only auto or WIDTHxHEIGHT; 1K / 2K / 4K returns 400.
Long synchronous jobs return 200 first and keep the connection alive, so failures also come back as 200; check the error field.
2.6 emits more output tokens than 2.5 at low / medium: at 1024², low measured about 820 and medium about 1,450; high and above match 2.5 (high about 1,760).

Build a requestJSON Schema

Pricing

Official price in USD. Actual charge = official price × group multiplier.

Input: $5 / 1M tokens
Cached input: $1.25 / 1M tokens
Output (incl. images): $30 / 1M tokens

Specs and limits

Source: spec/openai.json · spec/official.json, the same data the gateway validates against.

Size: Longest side ≤ 3840Sides in multiples of 16 · aspect ratio ≤ 3:1
Total pixels: 0.66–8.29 MP
Images per request: 1–10
Reference images: Up to 16Remote URLs ≤ 25 MiB each
Prompt: ≤ 32,000 characters

Endpoints

POST/v1/images/generations
POST/v1/images/edits
GET/v1/images/tasks?task_id={task_id}

Quick start

Replace sk-your-key with a key whose group includes this model.

curl https://api.hop-base.com/v1/images/generations \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.6-flare",
    "prompt": "A flat line icon of a paper plane",
    "output_format": "svg",
    "complexity": "low"
  }' \
  | jq -r '.data[0].b64_json' | base64 --decode > icon.svg

FAQ

How much does one image cost?: It is billed by tokens: $5 input, $1.25 cached input, and $30 output per 1M tokens, the same as 2.5. Output tokens grow with quality: at 1024², low uses about 820 (about $0.0246 per image) and high about 1,760 (about $0.0528); low / medium use more than 2.5, high and above the same. Actual charge = official price × group multiplier.
How do I get SVG or an animated GIF?: Set output_format to svg to get an SVG (n must be 1). Set it to gif for an animation: frame_count sets the number of frames (2–16, default 9) and frame_duration_ms the time per frame (20–2000 ms, default 200). For GIF, size is the size of one frame, so use auto or a small size such as 512x512: all frames are tiled into one sheet no larger than 3840 per side, and large frames fail.
What does complexity do?: It sets how much the model plans before rendering: low / medium / high, default high, with no separate fee. Any other value returns 400.
Can I edit by drawing annotations?: Yes. Send markup to /v1/images/edits: a transparent PNG the same size as the first input image with circles, arrows, or notes drawn on it (a multipart file part, or a URL or Data URL in JSON), and the model edits as annotated. markup cannot be combined with mask.

Same family

gpt-image-2.6-sunburst gpt-image-2.5-sunburst gpt-image-2.5-flare gpt-image-2