Text input
- Official OpenAI
- $5
- Here (−50%)
- $2.5
gpt-image-2.5-sunburst
GPT Image 2.5 Sunburst API — price per token
GPT Image 2.5 Sunburst is the higher-precision ChatGPT Images 2.5 model — slower than Flare, same official token rates as GPT Image 2, same Images HTTP conversion on this pool.
Pricing per 1M tokens
| Rate | Official OpenAI | Here (−50%) |
|---|---|---|
| Text input | $5 | $2.5 |
| Cached text input | $1.25 | $0.625 |
| Image input | $8 | $4 |
| Cached image input | $2 | $1 |
| Image output | $30 | $15 |
Cached text input
- Official OpenAI
- $1.25
- Here (−50%)
- $0.625
Image input
- Official OpenAI
- $8
- Here (−50%)
- $4
Cached image input
- Official OpenAI
- $2
- Here (−50%)
- $1
Image output
- Official OpenAI
- $30
- Here (−50%)
- $15
Every request is metered at the official rate first, then the flat 50% B2C discount is subtracted before it touches your prepaid balance. Context window: per request. Max output: 1–10 images (n).
Best for
- Higher-precision edits where Flare's faster everyday pass is not enough.
- Image editing against one to five PNG, JPEG or WebP references.
- Pipelines that already run GPT Image 2 here — Sunburst is a drop-in Images HTTP id.
Good to know
- gpt-image-2.5-sunburst is the alias of the immutable snapshot gpt-image-2.5-sunburst-2026-09-08 — the same model at the same price. Official token rates match GPT Image 2 and Flare.
- Generation is POST /v1/images/generations, editing is POST /v1/images/edits with up to five PNG, JPEG or WebP references (each ≤50 MB) — on the unified router (router.apitoken.sale/v1) or the legacy OpenAI host. Official GPT Image fields are accepted and converted. Multipart mask is ignored. Region inpaint (and hosted image_generation from a GPT text model) is POST /v1/responses with input_image plus tools image_generation; Chat Completions maps the same tool. Responses forwards output_format jpeg/webp and partial_images 1..=3, rewrites background=transparent to opaque, and drops input_fidelity. Mask image_url is PNG data URL only; no file_id.
- Images HTTP converts official client fields onto the Codex envelope (native size=auto, quality=low, background=opaque). Extra keys such as user, stream and moderation are ignored. quality medium/high/xhigh/max/auto still generate at low on a clean prompt. jpeg/webp transcode the native PNG locally. n=1..10 runs sequential native turns. size including 2048x2048 and 3840x2160 maps to 1:1 or 16:9; the PNG stays ~1.57 megapixels and the response size is the IHDR. background=transparent prepends a cutout sentence — real alpha is prompt-steered, not a JSON lock. Edits accept image or image[]; mask is ignored.
- Image output bills per image-output token; cached text/image input bills at 25% of the fresh rate. Production low canary (2026-09-09): HTTP 200, RGB PNG 1254×1254, 515 image-output tokens.
How to use GPT Image 2.5 Sunburst
Use gpt-image-2.5-sunburst with POST /v1/images/generations for new images or POST /v1/images/edits for reference edits at https://router.apitoken.sale/v1. Authenticate with Authorization: Bearer using the same key and prepaid balance. The default response contains PNG data in data[].b64_json. Top up your balance before sending API requests.
Frequently asked questions
How much does the GPT Image 2.5 Sunburst API cost?
Officially $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens (cached input at 25%). On apiToken.sale the same calls cost 50% less — $2.50/$4/$15 — at the flat discount applied to every call. Rates match GPT Image 2 and Flare.
What is the model ID for GPT Image 2.5 Sunburst?
gpt-image-2.5-sunburst, an alias of the immutable snapshot gpt-image-2.5-sunburst-2026-09-08. Send it as the model field of POST /v1/images/generations or /v1/images/edits on https://router.apitoken.sale/v1 (the legacy https://openai.api.apitoken.sale/v1 serves the same routes) with your Bearer key.
When should I pick Sunburst instead of Flare?
Sunburst is the higher-precision, slower Images 2.5 id. Flare is the faster everyday id. Both use the same Images HTTP conversion, the same five-leg tariff, and the same prepaid balance.
Can I send an OpenAI Images mask file?
POST /v1/images/edits ignores the mask field so the SDK still gets a full-image edit. To inpaint a region, call POST /v1/responses with a GPT text model, the source PNG as input_image, and image_generation.input_image_mask.image_url as a PNG data URL. You can set output_format jpeg|webp and partial_images 1..=3. file_id is not supported. Chat Completions maps the same tool onto Responses.
Does the same balance really cover image generation?
Yes. GPT Image 2.5 Sunburst debits the same prepaid balance as every Claude, GPT and Gemini model on the account — no separate image plan or key.
Run GPT Image 2.5 Sunburst on the OpenAI-compatible API at a flat 50% off — instant key, prepaid balance, card or crypto.