GPT-image โ
Request domain (international site): tokenhub-intl.tencentcloudmaas.com (Singapore) Endpoints: POST /v1/wand/og-image/generation (text-to-image), POST /v1/wand/og-image/edits (image editing)
GPT-image is a family of image generation models invoked synchronously through the TokenHub path, supporting both text-to-image and image editing. The result is already contained in the response body when the request returns โ no task polling required. Three versions are available โ 2.0, 2.5, and 2.6 โ all sharing the same endpoints and authentication, differing only in the model value.
All Three Versions Require Allowlisting
This family consists of test models and requires an allowlist entry granted by the backend team before it can be used. Contact the product team and submit your UIN and AppId. Allowlisting takes effect per model ID, so -f and -s must be requested separately; GIF output on 2.6 additionally requires the model's GIF preview entitlement.
TokenHub settles on an hourly cycle. Sales should apply for customer discounts in good time.
Choose a Version โ
The API reference is split per version. Click a card to see that version's model IDs, input parameters, request examples, and version-specific settings.
v2.6-f ยท v2.6-s- About 50% faster; Flare reaches ~6s at low tiers
- complexity decoupled from quality
- Animated GIF (2โ16 frames) and SVG vector output
- Markup editing, up to 16 reference images
v2.5-f ยท v2.5-s- Flare for speed / Sunburst for editing precision
- Adds the xhigh and max quality tiers
- Markup-based editing
- Lower per-image cost than 2.6 at low tiers
custom-model-og-v2- Single model ID, simplest parameter set
- Quality tiers up to high
- Text-to-image and image editing
- Supports mask-based inpainting
Version Comparison โ
| Item | 2.0 | 2.5 | 2.6 |
|---|---|---|---|
| Model ID | custom-model-og-v2 | custom-model-og-v2.5-fcustom-model-og-v2.5-s | custom-model-og-v2.6-fcustom-model-og-v2.6-s |
| Variant positioning | Single variant | Flare for speed / Sunburst for precision | Flare for speed / Sunburst for detail |
Quality tiers quality | auto / low / medium / high | Plus xhigh / max | Plus xhigh / max |
Planning tier complexity | โ | โ | โ
Three tiers, default high |
| Output formats | png / jpeg | png / jpeg | Plus webp / gif / svg |
mask inpainting | โ | โ | โ |
markup annotation editing | โ | โ | โ |
| Animated GIF | โ | โ | โ Text-to-image endpoint only |
| SVG vector | โ | โ | โ Text-to-image endpoint only |
| Generation speed | Baseline | Baseline | About 50% faster |
Selection Advice โ
- Prefer 2.6 for new integrations: at
highand above it costs the same as 2.5 but runs about 50% faster, so upgrading carries no cost penalty. - Stay on 2.5 if low-quality batch generation dominates: output image token counts at 2.6's
low/mediumtiers rise to roughly4.2ร/3.3ร, so per-image cost increases noticeably for icons, thumbnails, and bulk drafts. - GIF and SVG require 2.6, and both outputs are supported only on the text-to-image endpoint.
- For local edits only, pairing 2.6 with
complexity=lowskips re-laying-out the scene.
The Token Unit Price Is Identical Across All Three Versions
All are 1.6 USD per million tokens. Cost differences between versions come from the number of output image tokens, not the unit price. See "Billing and Pricing" in each version's document for the full conversion table.
Related Documents โ
| Document | Description |
|---|---|
| GPT-image 2.6 | Latest version: ~50% faster, complexity decoupled, GIF and SVG |
| GPT-image 2.5 | Two variants, xhigh / max tiers, markup-based editing |
| GPT-image 2.0 | Base version, with quality tiers up to high |
| Get an API Key | Applying for and managing TokenHub API Keys |