GPT-Image-2.5 Flare and Sunburst: GPT-Image-2 Rates, No Batch API
- What happened
- OpenAI launched ChatGPT Images 2.5 on September 8, 2026, and released two API models the same day: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst.
- Why it matters
- Both are listed at GPT-Image-2's standard rates ($8 per 1M image input tokens, $30 per 1M image output tokens), but neither supports the Batch API, and OpenAI says its GPT Image 2 calculator does not estimate 2.5 token use.
- What to do
- If you call GPT-Image-2 at standard rates, test Flare or Sunburst on your own prompts and measure tokens per image before switching. Keep Batch API jobs on GPT-Image-2: neither 2.5 model supports Batch.
Verdict: if you call GPT-Image-2 at standard API rates, test GPT-Image-2.5 Flare now. OpenAI lists identical token rates, so the rate card is not a reason to wait. If you run GPT-Image-2 through the Batch API, do not switch blind: neither 2.5 model supports the Batch API.
OpenAI published "Introducing ChatGPT Images 2.5" on September 8, 2026, saying it "helps turn your ideas, sketches, and reference photos into more personalized, polished images". The same day it released two API models, GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Neither API model is in our directory.
What happened
- Release: OpenAI's API changelog lists GPT Image 2.5 Sunburst and GPT Image 2.5 Flare as released on September 8, 2026. The model pages name the default snapshots:
gpt-image-2.5-flare-2026-09-08andgpt-image-2.5-sunburst-2026-09-08. - Positioning: OpenAI's image generation guide says to "Choose Sunburst for workflows where editing precision matters most, and Flare for fast, high-quality everyday image generation."
- Pricing: both model pages state "Token rates match GPT Image 2".
- Endpoints: both model pages list only the Image generation (
v1/images/generations) and Image edit (v1/images/edits) endpoints, and say the models do not support Batch. - Token estimates: OpenAI notes its GPT Image 2 calculator does not estimate GPT Image 2.5 token consumption.
We have not verified the ChatGPT rollout scope or OpenAI's speed claims, so this story covers the API release only.
Why it matters
At standard rates, the price does not separate the new models from the old one. OpenAI's pricing page, per 1M tokens:
| Model | Modality | Input | Cached input | Output |
|---|---|---|---|---|
| gpt-image-2.5-flare | Image | $8.00 | $2.00 | $30.00 |
| gpt-image-2.5-flare | Text | $5.00 | $1.25 | - |
| gpt-image-2.5-sunburst | Image | $8.00 | $2.00 | $30.00 |
| gpt-image-2.5-sunburst | Text | $5.00 | $1.25 | - |
| gpt-image-2 | Image | $8.00 | $2.00 | $30.00 |
| gpt-image-2 | Text | $5.00 | $1.25 | - |
Batch is where the models diverge. GPT-Image-2 has a Batch rate of $4.00 per 1M image input tokens ($1.00 cached) and $15.00 per 1M image output tokens, with text input at $2.50 ($0.625 cached). Neither 2.5 model supports the Batch API, per OpenAI's model pages. A Batch pipeline that switches to 2.5 has to move to synchronous Image API calls at the standard rate: double GPT-Image-2's Batch rate per token.
Per-token parity is not per-image parity. What one image costs depends on how many tokens it consumes. OpenAI's GPT Image 2 calculator does not estimate 2.5 usage, and we have not measured it. Run your own prompts before you assume a swap costs the same per image.
What changes for you
| If you... | Do this |
|---|---|
| Call GPT-Image-2 at standard rates | Test Flare on your own prompts. Compare output quality and tokens per image before switching |
| Run GPT-Image-2 through the Batch API | Stay on GPT-Image-2 for Batch jobs. Neither 2.5 model supports Batch, so switching means synchronous calls at the standard rate |
| Edit images where precision matters | Test Sunburst, which OpenAI positions for workflows where editing precision matters most |
FAQ
What do GPT-Image-2.5 Flare and Sunburst cost on the API? Both are listed at $8.00 per 1M image input tokens, $2.00 per 1M cached image input tokens and $30.00 per 1M image output tokens, with text input at $5.00 per 1M ($1.25 cached). Those are the same standard rates OpenAI lists for GPT-Image-2.
Is there a Batch discount for the 2.5 models? No. Neither model supports the Batch API, per OpenAI's model pages as of September 30, 2026. GPT-Image-2 lists Batch rates at half its standard price.
Should I use Flare or Sunburst? OpenAI points Flare at fast, high-quality everyday image generation and Sunburst at workflows where editing precision matters most. The token rate is the same, so choose on output quality and speed on your own prompts.
When were the API models released? September 8, 2026, per OpenAI's API changelog.
What to do
- 1 If you call GPT-Image-2 at standard rates, test GPT-Image-2.5 Flare on your own prompts and compare output quality and tokens per image, not the rate card.
- 2 If you run GPT-Image-2 through the Batch API, keep those jobs on GPT-Image-2: neither Flare nor Sunburst supports Batch, so switching means synchronous calls at the standard rate.
- 3 Test GPT-Image-2.5 Sunburst for image editing where precision matters most.
Affected tools & models
Never need to catch up again
The weekly delta — only verdict changes and act-now items. No digest filler.