Two model IDs
Sunburst and Flare are separate identifiers on the gateway, each with its own rate line, so either can be used without the other.
What GPT Image 2.6 is
The next release in OpenAI's GPT Image line, announced on November 4, 2026 and enabled on TokenRa the following day. It ships in two variant tiers, Sunburst and Flare — the same structure as 2.5 — and both run through the OpenAI-compatible image endpoint. Each tier has its own model identifier and its own line on the bill, so you can route different workloads to different tiers without touching the rest of your integration.
Live: both variants enabled
gpt-image-2.6-flare and gpt-image-2.6-sunburst are
callable right now at $0.012 per request. Current rates always live on the
pricing page.
Same request surface
The endpoint, headers and parameters match 2.5: model, prompt, and the optional size, quality and n settings. Moving generation is a model-string change.
Lower unit cost
$0.012 per request is 40% below the $0.02 charged for 2.5, on the same per-request billing unit.
Sunburst and Flare: the two variants
Sunburst and Flare are the two tiers released with this generation. On TokenRa each is a distinct identifier, enabled and priced on its own — 2.5 already works this way, with a separate Sunburst and Flare identifier. Treat them as two models to test, not as one model with a quality slider.
Sunburst tier
Model ID gpt-image-2.6-sunburst, billed at $0.012 per request.
Flare tier
Model ID gpt-image-2.6-flare, billed at $0.012 per request.
Test both, then decide
Run the same prompt set through both and pick on output quality, latency and cost per accepted image — not on the tier name. Both cost the same per request.
Both identifiers are live
No waitlist and no staged rollout: both strings resolve on the image endpoint today. If a client caches model lists, confirm them in the AI model catalog.
GPT Image 2.6 vs GPT Image 2.5
The differences that matter are price and generation. Both are enabled, both bill per request, both take the same request body. The table reflects the live gateway as of November 5, 2026.
| GPT Image 2.6 | GPT Image 2.5 | |
|---|---|---|
| Announced | November 4, 2026 | Earlier release |
| Variants | Sunburst, Flare | Sunburst, Flare |
| Status on TokenRa | Enabled (November 5, 2026) | Enabled |
| Model ID | gpt-image-2.6-flare / gpt-image-2.6-sunburst |
gpt-image-2.5-flare / gpt-image-2.5-sunburst |
| TokenRa price | $0.012 per request | $0.02 per request |
| vs 2.5 | 40% lower per request | Baseline |
| Choose it when | You want a lower cost per image, or the newest generation | You already validated output and do not want to re-test |
Rates are the published gateway prices. Per-request billing means retries and rejected images each cost a request, so compare cost per accepted image, not per call.
This is not Microsoft MAI-Image-2.6
Search interest in "gpt image 2.6" and "mai image 2.6" rose in the same week, and they are two different models from two different companies. This page is OpenAI's image model. MAI-Image-2.6 is a Microsoft image model published on microsoft.ai and documented separately. Check the identifier prefix before you integrate anything.
- Who makes it: this line — OpenAI. MAI-Image-2.6 — Microsoft.
- Identifier shape: the OpenAI line uses a
gpt-image-*prefix; Microsoft's MAI models are a separate family with their own naming. - Where the docs live: OpenAI's model documentation for the GPT Image line, microsoft.ai for MAI-Image. Do not mix release notes between the two.
- Before you switch: quality claims in either vendor's launch material are vendor-specific. Compare both on your own prompts and assets.
Availability and how to get access
Nothing special is required: an account and an API key get you both tiers immediately, with no application and no waitlist. The only account floor is the $10 top-up minimum, and a zero-balance account is rejected before the model is called.
Create a TokenRa account
Register, then open the dashboard.
Create your API key
Open the API key section, create a key, and store it server-side.
Send a request with a 2.6 ID
Point the model field at flare or sunburst and run a representative
prompt set before routing production traffic.
Keep the model ID in configuration
Hard-coding a model string turns a tier change into a code change. Read the identifier from environment configuration so moving between flare, sunburst and 2.5 is a one-line switch, and log which identifier produced each output.
How to call GPT Image 2.6
Swap the model string into the request you already send for 2.5. The endpoint, headers and billing unit are unchanged. For complete API documentation, see the GPT Image 2.6 Flare API docs and GPT Image 2.6 Sunburst API docs.
curl https://tokenra.io/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-image-2.6-flare","prompt":"A clean editorial illustration of a sunrise over a city"}'
For the other tier, send "model":"gpt-image-2.6-sunburst" with the same
body.
- Keep the key on the server. Never ship a TokenRa key in browser code.
- Set the client timeout above 120 seconds: images are returned base64-encoded by default and a single call takes about 30 seconds in testing.
- Log the model identifier alongside the request so flare and sunburst output stay comparable on the same prompt set.
Pricing
Both tiers are $0.012 per image request — a flat charge per call, not a per-token charge. GPT Image 2.5 is $0.02 per request, so this generation is 40% cheaper at the unit level. That is the whole pricing change: same billing unit, lower rate, either variant.
$0.012 / request
Flat rate per image generation request, for either tier.
40% lower
Against the $0.02 per request charged for 2.5 on the same endpoint.
Per request
Request count is the billing unit on the image endpoint, not tokens.
OpenAI model
GPT Image 2.6 is developed by OpenAI. TokenRa provides API access as a routing gateway and is not the model developer or owner.
Data handling
Prompts, input images, and generated output are processed under the applicable provider and TokenRa terms. Review those terms and the privacy policy before sending sensitive or regulated content, and confirm output rights for your use case.
Production checklist before switching to 2.6
- Keep the model identifier in configuration, not hard-coded, so a tier change is a config change.
- Re-run your prompt and asset set on 2.6 and compare against 2.5 output before routing traffic to it.
- Canary a small share of requests first, and keep a fallback to
gpt-image-2.5-flare. - Set client timeouts above 120 seconds and budget request counts, not tokens.
- Log the identifier and the outcome of each call so cost per accepted image stays measurable.
- Review data handling and output rights for your use case before launch.
FAQ
Is GPT Image 2.6 available on TokenRa?
Yes. Both variants are enabled on the image endpoint at $0.012 per request.
What is the model ID?
gpt-image-2.6-flare and gpt-image-2.6-sunburst. Confirm both in the catalog before shipping.
How much does it cost?
$0.012 per image request, billed per request rather than per token. GPT Image 2.5 is $0.02, so 2.6 is 40% lower.
When was it released?
OpenAI announced GPT Image 2.6 on November 4, 2026, and TokenRa enabled both variants on November 5, 2026.
Is 2.6 cheaper than 2.5?
$0.012 versus $0.02 per request — 40% lower. Compare per accepted image rather than per request, because retries and rejected images each cost a request.
Is this the same model as Microsoft MAI-Image-2.6?
No. GPT Image 2.6 is an OpenAI image model. MAI-Image-2.6 is a Microsoft image model documented on microsoft.ai. They are different models from different vendors, with different identifiers and different documentation.
What is the difference between Sunburst and Flare?
They are two variant tiers, each a separate model identifier with its own output character. Both are billed at $0.012 per request on TokenRa, so test both on your own prompts rather than picking on the tier name.
Can I use it in production today?
Yes — call either identifier with a server-side key and log the identifier with each request.
Which endpoint do I call?
Send requests to https://tokenra.io/v1/images/generations with a server-side TokenRa API key.
"GPT Image", "OpenAI", and related names may be trademarks of their respective rights holders. TokenRa is an independent API gateway and this page is not an official vendor page or endorsement. Microsoft and MAI-Image are trademarks of Microsoft Corporation.
Last updated: November 5, 2026