Read This First: GPT Image 1 Is Being Retired
If you're integrating the GPT Image 1 API today, the most important fact is the calendar: OpenAI's deprecations page lists gpt-image-1 for shutdown on October 23, 2026. (DALL·E 2 and 3 were already removed from the API on May 12, 2026.) So while GPT Image 1 still works, any new build should target its successor, GPT Image 2 — OpenAI's current flagship image model. This guide covers GPT Image 1's pricing, how it stacks up, and the migration.
GPT Image 1 Pricing (Legacy)
GPT Image 1 uses token-based billing. In practice it runs roughly $0.011–$0.25 per image depending on quality and resolution — with a cheaper gpt-image-1-mini ($0.005–$0.05) for drafts. There's also a $0.009–$0.20) that some teams pin for stability. These remain callable in-flight, but they're on the way out.gpt-image-1.5 interim flagship (
Why GPT Image 2 Is the Upgrade
GPT Image 2 (released April 21, 2026) is the first reasoning image model — it plans composition and text placement before rendering, giving near-perfect text and reliable complex layouts. Verified OpenAI rates: $30/1M output tokens, which works out to ~$0.05 per medium 1024px image, ~$0.21 at high quality, up to ~$0.35 at 4K. Batch API halves it. Full breakdown in the GPT Image 2 pricing guide.
| GPT Image 1 (legacy) | GPT Image 2 (current) | |
|---|---|---|
| Status | Shuts down Oct 23, 2026 | Active flagship |
| Text rendering | Good | Near-perfect (reasoning) |
| Max resolution | 1024px tiers | 2K (4K via variants) |
| Per image | ~$0.011–$0.25 | ~$0.05 (med) / ~$0.21 (high) |
| Editing | Basic | High-fidelity, instruction-based |
Note: there is no gpt-image-1 on gateways like LinkModel — the catalog carries the current gpt-image-2, which is the right migration target anyway.
What About "GPT-4o Image Generation"?
Some searches look for a "GPT-4o image API." In practice, OpenAI's image generation now lives in the dedicated GPT Image models, not GPT-4o. If that's what you're after, gpt-image-2 is the endpoint you want.
Migrating in Practice
The request shape is simple on a gateway — submit a prompt, poll for the image:
curl -X POST https://api.linkmodel.ai/api/v1/image-generation \
-H "Authorization: Bearer $LINKMODEL_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "model": "gpt-image-2", "prompt": "Product label reading \"COLD BREW\" in bold, minimalist, 2K" }'Migration checklist:
- Find every
gpt-image-1/gpt-image-1-mini/ DALL·E reference in your code. - Swap to
gpt-image-2(and update any resolution/quality params). - Re-test prompts — GPT Image 2 rewards more specific, structured prompts; see the prompting guide.
- Finish before October 23, 2026.
Cheapest Way to Run It
On LinkModel, GPT Image 2 is available up to 30% below OpenAI's rate with fixed per-request pricing and one API key across image, video, and chat models. For the alternatives (cheaper bulk models, Google's Nano Banana), see best AI image generation APIs and Nano Banana Pro vs GPT Image 2.
Start free with a $1 credit and validate GPT Image 2 on your prompts before you migrate.
