Compare GPT Image Models
GPT Image 2.5 Flare and Sunburst are available on EvoLink alongside GPT Image 2 and 1.5. Choose a starting model for everyday generation or precise editing, then open its pricing, Playground, and API reference.
Everyday generation
GPT Image 2.5 Flare
Editing precision
GPT Image 2.5 Sunburst
Existing workflow baselines
GPT Image 2 / 1.5
Compare the GPT Image models
Start with workflow fit. The 1024×1024 amounts below are output-cost references, not complete request quotes or measured cost per accepted image.
| Model | Best for | 1024x1024 low | 1024x1024 medium | 1024x1024 high | Notes |
|---|---|---|---|---|---|
GPT Image 2.5 Flare Everyday generation | UI drafts, campaign variations, and frequent generation with explicit acceptance rules. | $0.006 | $0.013 | $0.053 | Available on EvoLink. OpenAI positions Flare for everyday generation; measure actual usage and acceptance on your task. |
GPT Image 2.5 Sunburst Editing precision | Reference-based product edits and multi-step revisions where protected details matter. | $0.006 | $0.013 | $0.053 | Available on EvoLink. OpenAI prioritizes editing precision with longer generation times; test the full edit sequence. |
GPT Image 2 Previous generation | Existing GPT Image 2 workflows and paired migration baselines. | $0.006 | $0.053 | $0.211 | Retain a working configuration as a baseline while testing a 2.5 replacement. Check this model’s own parameters and current pricing. |
GPT Image 1.5 Existing workflows | Teams with prompts and assets already validated on GPT Image 1.5. | $0.009 | $0.034 | $0.133 | Retain a working configuration as a baseline while testing a 2.5 replacement. Check this model’s own parameters and current pricing. |
Output-cost references reuse the site pricing catalog; they are not complete EvoLink request quotes. Inputs and actual usage also affect charges. Verify current rates on the model pages.
How to decide which GPT Image model to use
Follow these 4 rules to narrow down your choice.
Start with the task
For everyday generation and rapid iteration, start by testing Flare. OpenAI positions it as the default for most applications.
Test the details you must preserve
For product references, masked edits, and multi-step revisions, compare Sunburst with Flare using the same acceptance rules.
Keep an existing baseline
If GPT Image 2 or 1.5 already works for your workload, preserve that configuration until a replacement passes a paired test.
Compare delivery cost
Use actual usage, retries, acceptance, and review time. A higher quality label does not guarantee a deliverable, and shared token rates do not guarantee equal bills.
If you already know your use case, find the recommended starting point in the table below.
Choose a GPT Image model by workflow: generation, editing, and creative production
Match your primary image task to the right model.
| Your task | Recommended start | Good fit if... | Watch out for |
|---|---|---|---|
| Everyday generation and UI drafts | GPT Image 2.5 Flare | You need frequent variations with clear acceptance checks | Verify labels, layout, and cost across the full batch |
| Precise editing and product revisions | GPT Image 2.5 Sunburst | Preserving approved details through edits matters most | Test the full sequence; longer waits must still meet deadlines |
| Existing GPT Image 2 workflows | GPT Image 2 + a 2.5 comparison | Your current configuration already meets delivery requirements | Keep it as a baseline until migration improves the relevant task |
| Existing GPT Image 1.5 workflows | GPT Image 1.5 + a 2.5 comparison | Your prompts and assets are already validated on 1.5 | Recheck parameters and outputs instead of assuming a drop-in upgrade |
GPT Image workflows: generation, editing, product visuals, and batch creative production
See how GPT Image models fit into real image generation and editing pipelines.
Campaign assets and UI drafts
Use Flare for an initial evaluation of layouts and creative variations. Check required text and regions, then compare the cost of accepted outputs.
Reference and masked editing
Use Sunburst as a candidate when editing precision is the main constraint. Inspect protected details and untouched regions after each change.
Product visuals and multi-step revisions
Compare both variants on the same product references. Reject altered labels or silhouettes; composite the original product if identical pixels are required.
Production migration
Retain the GPT Image 2 or 1.5 configuration that already works. Adopt a 2.5 configuration only after fresh validation of acceptance, deadlines, and costs.
View GPT Image model details
Pick a model, then visit its detail page for capabilities, pricing, and integration.
GPT Image 2.5 Flare
Everyday generation
- Output-cost reference
- $0.006 / $0.013 / $0.053
GPT Image 2.5 Sunburst
Editing precision
- Output-cost reference
- $0.006 / $0.013 / $0.053
GPT Image 2
Previous generation
- Output-cost reference
- $0.006 / $0.053 / $0.211
GPT Image 1.5
Existing workflows
- Output-cost reference
- $0.009 / $0.034 / $0.133
Access GPT Image models through one EvoLink API
Access the four models listed here with one EvoLink account and API key. Use the exact model ID, check each route’s supported parameters, and compare usage in one billing system.
For 2.5, select model="gpt-image-2.5-flare" or model="gpt-image-2.5-sunburst". EvoLink quality defaults to medium; there is no gpt-image-2.5 alias.How to think about GPT Image cost: quality, resolution, and volume
Choose a tested quality setting
EvoLink 2.5 exposes low, medium, high, xhigh, and max, with medium as the default. Use the lowest-cost tested configuration that passes your task.
Include inputs and editing steps
Output size and quality affect the estimate. Reference images, prompt tokens, repeated edits, and retries can add to the complete bill.
Compare accepted deliverables
Divide reconciled batch charges by accepted final images. Track review time and missed deadlines separately; identical token rates do not settle the choice.
1024×1024 output-cost references
These references reuse the existing pricing catalog for low, medium, and high. GPT Image models are token-billed; inputs, extra attempts, and actual usage can change the total. Open a model page for the current EvoLink rates and calculator.
GPT Image 2.5 Flare
$0.006 low
$0.013 medium
$0.053 high
Available on EvoLink. OpenAI positions Flare for everyday generation; measure actual usage and acceptance on your task.
GPT Image 2.5 Sunburst
$0.006 low
$0.013 medium
$0.053 high
Available on EvoLink. OpenAI prioritizes editing precision with longer generation times; test the full edit sequence.
GPT Image 2
$0.006 low
$0.053 medium
$0.211 high
Retain a working configuration as a baseline while testing a 2.5 replacement. Check this model’s own parameters and current pricing.
GPT Image 1.5
$0.009 low
$0.034 medium
$0.133 high
Retain a working configuration as a baseline while testing a 2.5 replacement. Check this model’s own parameters and current pricing.
Related GPT Image guides
Use the family page for model selection. Use the guides below for product access, release-status clarity, and direct head-to-head comparisons.
Flare vs Sunburst: which should you choose?
Use six workloads, two task briefs, and a cost worksheet to choose between the two 2.5 variants.
GPT Image 2.5 release date and status
Use this page for what OpenAI released on September 8, 2026 and what changed from GPT Image 2.
GPT Image 2 vs GPT Image 1.5
Use this comparison when your real decision is whether to move to GPT Image 2 now or keep GPT Image 1.5 as your production default.
GPT Image 2 vs Nano Banana 2
Use this comparison when your shortlist includes GPT Image 2 and Google's Nano Banana 2.
Mona Lisa 1 Arena test
Use this page for the first-hand test of the anonymous Arena checkpoint and the evidence boundary between it and any official OpenAI model.
GPT Image Family FAQ
Common questions about choosing a GPT Image model, access, and billing.