GPT-Image-2-VIP Image Gen/Editing

GPT image generation reverse-engineered model gpt-image-2-vip (Codex line). Flat $0.03/image. Supports 30 explicit sizes (10 ratios × 3 resolution tiers: 1K / 2K / 4K). Same call format as gpt-image-2-all. About 90–150s per image — for workloads that need locked output dimensions.

Overview

gpt-image-2-vip is the GPT image generation reverse-engineered model on the Codex line, available on the API易 platform. Same flat $0.03/image as gpt-image-2-all and identical request/response format — the only meaningful difference is that vip accepts a size field with 30 common sizes (10 aspect ratios × 3 resolution tiers: 1K Fast / 2K Recommended / 4K Detail), including 4K.

🎨 Positioning: use gpt-image-2-vip when you need to lock the output size (e-commerce hero shots, poster templates, video thumbnails, 4K wallpapers, etc.). Just swap the model field to gpt-image-2-vip and add a size field — every other line of code stays identical to gpt-image-2-all.

Chat API

OpenAI Chat Completions format — one endpoint for both text-to-image and reference-image editing, accepts online image URLs directly.

Text-to-Image API

/v1/images/generations — text prompt + size for explicit output dimensions.

Image Editing API

/v1/images/edits — multipart upload with edit/fusion instructions.

Key differences vs gpt-image-2-all

gpt-image-2-vip and gpt-image-2-all are both reverse-engineered channels, same price, same call code. They mirror each other — swap the model field on the same request and behavior is largely identical. The differences:

Dimension	gpt-image-2-all	gpt-image-2-vip
Channel	Reverse-engineered ChatGPT web	Reverse-engineered Codex line
Price	$0.03 / image	$0.03 / image (flat across all sizes)
size parameter	❌ Not accepted (describe in prompt)	✅ 30 sizes incl. 4K
4K (e.g. 3840x2160)	❌	✅ 4K Detail tier
Generation time	~30–60 seconds	~90–150 seconds (on par with the official gpt-image-2)
quality parameter	❌ Not accepted	❌ Not accepted (do not pass)
Endpoints	/chat/completions + /images/generations + /images/edits	Same as left (identical)
Response format	url / b64_json (already prefixed)	Same as left
Best for	Prompt-driven, size-insensitive	Need locked output size (incl. 4K)

One-line decision: don’t need strict size, want fastest output → gpt-image-2-all; need locked size or 4K → gpt-image-2-vip; need a quality knob or strict OpenAI-API field parity → use the official gpt-image-2.

Core Features

Locked output size

The size field accepts 30 common sizes — e-commerce hero shots, poster templates, 4K wallpapers all output at exact pixels.

4K High Resolution

The 4K Detail tier covers 2880×2880 / 3840×2160 / 3840×1632 etc., suitable for large deliverables.

Flat pricing across all sizes

1K / 2K / 4K all cost $0.03/image — no surcharge for 4K.

Same call format as -all

Request structure, fields, and response shape are identical to gpt-image-2-all — switch models with just the model string.

High Text Rendering

Stable rendering of Chinese/English text, signs, and poster text — ideal for infographics and marketing assets

Chinese Prompt Friendly

Native understanding of Chinese descriptions without translation

Natural-Language Editing

Edit via conversational descriptions, no masks required, supports multi-turn iteration

Triple Endpoint Support

Compatible with /images/generations, /images/edits, and /chat/completions

Pricing

Model	Billing	Price	Output
gpt-image-2-vip	Per-call	$0.03 / image	1 image per call, size field locks output dimension

Billing notes:

• Flat $0.03/image across all 30 sizes — no surcharge for 4K Detail
• Failed requests are not charged (auth failures, parameter validation errors)
• For N images, call the API N times in parallel

Group Setup

gpt-image-2-vip lives on the Default group — no extra group needed. The reverse channel currently has stable supply, so there’s no enterprise-group fallback story like the official-relay gpt-image-2 has.

Model	Group	Notes
gpt-image-2-vip	Default	Codex reverse line, flat $0.03/img, ~90–150s

Advanced (when you also use gpt-image-2-all and the official-relay gpt-image-2): if your token covers all three models, set the token’s group priority like this:

• First priority: image2Enterprise (1.2x enterprise group, dedicated stable lane for the official relay)
• Default fallback: Default (both reverse models live here and route by model)

Result: official-relay gpt-image-2 rides the enterprise lane for stability, while the two reverse models stay on the default group — one token covers all three, no interference.

📖 About the image2Enterprise group: /en/live/2026-04/image2-enterprise-stable

Technical Specs

Attribute	Value
Model name	gpt-image-2-vip
Channel type	Official reverse-engineered (Codex line)
Pricing	$0.03 / image, per-call (flat across all sizes)
Generation time	~90–150 seconds (on par with the official gpt-image-2; slower than gpt-image-2-all’s 30–60s)
size parameter	✅ 30 sizes: 10 ratios × 3 resolution tiers (1K Fast / 2K Recommended / 4K Detail)
4K support	✅ 4K Detail tier (e.g., 3840x2160 / 2880x2880)
quality parameter	❌ Not supported, do not pass
n parameter	❌ Not supported, single image per call
Default response format	url (R2 CDN accelerated link, ~1-day validity)
Alternative format	b64_json (already prefixed with data:image/png;base64,)
Chinese prompts	✅ Natively supported
Capabilities	Text-to-image, single-image editing, multi-image fusion, natural-language editing (all three endpoints)

⏰ Image URL validity: ~1 day (default)

The default url field is an R2 CDN link that expires in about 24 hours — requests after that will 404. For images that need long-term retention, download and persist them to your own storage as soon as possible after generation, or use the b64_json response format.

Endpoints

gpt-image-2-vip is compatible with the exact same three endpoints as gpt-image-2-all. Just swap the model field and add a size if needed:

Endpoint	Purpose	Content-Type	Best for
POST /v1/chat/completions	Chat-based (text-to-image / editing / multi-turn / reference images)	application/json	Pass online image URLs directly; one endpoint for both generation and editing
POST /v1/images/generations	Text-to-image	application/json	OpenAI Images API standard format — same code can hit both official and reverse channels
POST /v1/images/edits	Image editing (single/multi)	multipart/form-data	OpenAI Images API standard format — same code can hit both official and reverse channels

Domain options: api.apiyi.com is the main domain. You can also use alternate gateway domains such as b.apiyi.com / vip.apiyi.com. Response behavior is identical.

Supported sizes (full 30-size table)

gpt-image-2-vip supports 10 aspect ratios × 3 resolution tiers = 30 sizes. Pass size: "WIDTHxHEIGHT" (lowercase ASCII x) directly in the request body.

1K Fast — drafts and low-cost iterations

Ratio	Name	Pixels
1:1	Square	1280x1280
2:3	Portrait	848x1280
3:2	Photo	1280x848
3:4	Portrait	960x1280
4:3	Standard	1280x960
4:5	Social	1024x1280
5:4	Large	1280x1024
9:16	Story	720x1280
16:9	Wide	1280x720
21:9	Cinema	1280x544

2K Recommended — default tier (most production outputs)

Ratio	Name	Pixels
1:1	Square	2048x2048
2:3	Portrait	1360x2048
3:2	Photo	2048x1360
3:4	Portrait	1536x2048
4:3	Standard	2048x1536
4:5	Social	1632x2048
5:4	Large	2048x1632
9:16	Story	1152x2048
16:9	Wide	2048x1152
21:9	Cinema	2048x864

4K Detail — large deliverables

Ratio	Name	Pixels
1:1	Square	2880x2880
2:3	Portrait	2336x3520
3:2	Photo	3520x2336
3:4	Portrait	2480x3312
4:3	Standard	3312x2480
4:5	Social	2560x3216
5:4	Large	3216x2560
9:16	Story	2160x3840
16:9	Wide	3840x2160
21:9	Cinema	3840x1632

Flat pricing across all 30 sizes: $0.03/image. No surcharge for 4K Detail.

Picking a tier:

• 1K Fast — drafts, thumbnails, A/B tests. Fastest output (price is flat, but iteration loop is shorter).
• 2K Recommended — default tier. Covers most production outputs (e-commerce hero shots, posters, infographics).
• 4K Detail — print, large displays, video thumbnails, desktop / outdoor large format.

Minimal call example (only pass size, do not pass quality):

curl "https://api.apiyi.com/v1/images/generations" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $YI_API_KEY" \
  -d '{
    "model": "gpt-image-2-vip",
    "prompt": "Product shot of a white ceramic mug on a gray desk, soft natural light, clean background",
    "size": "2048x1360"
  }'

Best Practices

1. Compress input images to under 1.5MB (image edit / multi-image fusion)

   Compress each image you upload to under 1.5MB (JPEG quality 80-90 / down-sized resolution); apply the same cap per image in multi-image fusion. Sporadic shell_api_error / Unknown error responses are most often triggered by oversized inputs — compressing measurably improves success rate and latency. Output resolution is governed by the size field, not by input size — shrinking the input only speeds things up, it does not hurt quality. Stuffing 4K / 8K into the prompt does not produce a 4K image; resolution is set by size, not by prompt fluff.

2. Pick the size tier by deliverable

   1K Fast for drafts, 2K Recommended for production, 4K Detail for print/large displays. Pricing is flat — pick by need.

3. Use lowercase ASCII x in size

   Send "size": "1536x1024" — not 1536×1024, not uppercase X.

4. Do not pass quality or n

   quality is rejected; n returns 1 image per call regardless — for multiple images, call in parallel.

5. Use a 300s timeout

   Typical generation is 90–150s, but image upload / download time and peak-tail latency push it higher. Set 300s as a conservative baseline.

6. Choose response format by need

   Use b64_json for direct web rendering; url for server-side storage/forwarding.

7. Share code with -all

   Same code works for both — switch model between gpt-image-2-all and gpt-image-2-vip as needed. Use vip when you need locked size, switch back to -all for fastest iteration.

Error Codes and Retries

Status	Meaning	Suggestion
400	size not in the 30-size set, or malformed	Use the exact strings from the table above
401	Invalid token	Check Bearer Token
429	Rate limit / quota exhausted	Exponential backoff retry
500 (4K sporadic)	OpenAI upstream compute fluctuation; 4K Detail tier hits this more often	Drop to 2K Recommended and retry; if 4K is mandatory, switch to official-proxy gpt-image-2 + image2Enterprise group
5xx (other)	Transient gateway/backend error	Retry 1–2 times
Timeout	Codex peak + 4K long tail	Set client timeout ≥ 300s (conservative)

Client recommendations:

• Request timeout starting at 300 seconds (conservative; typical 90–150s, but 4K Detail + peak tails go higher)
• Use exponential backoff for 5xx and timeouts (2–3 retries recommended)
• Log the request-id response header for debugging

FAQ

Can I share code between vip and -all?

Yes, almost identical. All three endpoints (/v1/chat/completions, /v1/images/generations, /v1/images/edits) share request fields, response fields, and b64_json prefix behavior. The only differences:

1. model field: gpt-image-2-vip ↔ gpt-image-2-all
2. size field: vip accepts the 30-size set; -all rejects size (size goes into the prompt instead)

Practical pattern: keep one codebase with an if model == 'vip': payload['size'] = ... switch.

Why is vip so much slower?

gpt-image-2-vip uses the Codex reverse channel — typical 90–150 seconds, on par with the official gpt-image-2 (100–120s) and slower than ChatGPT-web-line gpt-image-2-all (30–60s). For latency-sensitive workloads, prefer gpt-image-2-all; switch to vip only when you need locked size or 4K.

Does size have to be exactly from the table? What if I send 1024x768?

Yes — stick to the 30-size set. Off-list sizes may trigger upstream invalid_request_error. Pick the closest tier for your deliverable.

Why does 4K frequently return 500? How do I get reliable 4K?

Symptom: at the 4K Detail tier (e.g., 3840x2160 / 2880x2880), status_code: 500 errors are easier to trigger, with upstream returning invalid_request_error:

{
  "status_code": 500,
  "error": {
    "message": "An error occurred while processing your request. ... Please include the request ID xxxxxxxx in your message.",
    "type": "invalid_request_error",
    "code": null
  }
}

Root cause: OpenAI compute fluctuation — not your request parameters. The same payload usually goes through at 2K. The Codex reverse channel is more sensitive to large outputs like 4K, especially at peak hours.

Mitigation (by cost-effectiveness):

1. Prefer 2K Recommended (e.g., 2048x1360 / 2048x2048) — significantly higher success rate, same $0.03/image
2. Send fewer input images for img2img / multi-image fusion — the Codex reverse channel struggles under heavy input load, further pushing up 4K failure rates; pre-compressing each input image under 1.5MB also helps
3. For guaranteed 4K — switch to the official-proxy gpt-image-2 + image2Enterprise group. The official-proxy 4K is pricier (~$0.3+/image), but markedly more stable — appropriate when 4K delivery is a hard requirement.

📖 Field note: /en/live/2026-05/gpt-image-2-vip-4k-tips

Should I compress input images? Does writing 4K / 8K in the prompt help?

Yes, strongly recommended. Compress each input image to under 1.5MB (JPEG quality 80-90 / down-sized resolution): sporadic shell_api_error / Unknown error responses are most often triggered by oversized inputs, and compressing measurably improves success rate and latency. Note: 1.5MB is the recommended ceiling for reliability and speed; the 10MB number in the FAQ above is the gateway hard limit.

Don’t worry about compression hurting quality — output resolution is governed by the size parameter, not by your input size. Shrinking the input only speeds things up.

Stuffing 4K / 8K into the prompt does not actually produce 4K output. If your prompt says 8K ultra HD but you set size to 1024x1024, you still get a 1K-quality image. For 4K, set it in the size field — 1K / 2K / 4K all cost the same flat $0.03/image across the 30-size set.

📖 Source: /en/live/2026-05/gpt-image-2-vip-unknown-error

Is 4K really not surcharged?

No surcharge. The 4K Detail tier (3840x2160 / 2880x2880 etc.) costs the same $0.03/image as 1K and 2K.

Does it support n? What happens if I pass n=3?

No. This model returns 1 image per call — for multiple images, use repeated / concurrent calls instead.

⚠️ Important: if you pass n=3 in the request, billing will be 0.03 × 3 = $0.09, but only 1 image is actually returned. Drop the n field to avoid wasted charges.

If the content is rejected or the model replies 'I can't do that', is it billed?

This is a reverse-engineered channel using synchronous chat-style responses. Outcomes split into two cases with different billing rules:

1) HTTP 5xx returned → NOT billed

When upstream content policy hard-blocks the request, you’ll see something like:

{
  "error": {
    "message": "Image was not generated as expected. Please adjust the prompt and retry (traceid: 0672821c6951af183dbf847130caaf16)",
    "localized_message": "Unknown error",
    "type": "invalid_request_error",
    "param": "",
    "code": null
  }
}

These hard errors are not billed. Ask the user to adjust the prompt and retry.

2) HTTP 200 with a text “soft refusal” → BILLED

When the model soft-refuses inside the conversation (e.g. “I can’t do that”, “Sorry, this request involves…”), at the protocol level it looks like a normal chat completion, so it is billed. The reverse channel cannot reliably distinguish “refusal text” from “image output” at the protocol layer.

Why we can’t just waive soft refusals

Auto-waiving every soft refusal would mean the platform absorbs every failed upstream call. More importantly, frequently triggering upstream content safety also raises the risk of the supplier’s account being banned — that’s a real supply-side cost we can’t fully eliminate.

Recommendations for integrators

• ✅ Pre-filter and warn users: add a keyword/scenario filter at the frontend or gateway (real-person names, copyrighted characters, sensitive topics) and surface a UI hint like “Celebrity / IP topics may fail and still be billed by upstream policy.” This cuts wasted charges sharply.
• ✅ Monthly reimbursement for consumer products: we understand consumer-facing products can’t fully gate user input. If your monthly spend is large enough ($1000+/month), you can batch your logs monthly (short-latency calls are usually soft refusals) and contact support for a one-off manual credit — no need to file per-call appeals.

📖 Related: 500 errors are usually content-policy hits (not billed)

Do I need to add data:image/png;base64, prefix to b64_json?

No. The b64_json field already includes the prefix. You can use it directly as <img src> or write it to a file. If your code follows the old “prepend prefix” pattern, you’ll produce a broken data URL — add a startsWith('data:') check first.

What's the max reference image size and supported formats?

Recommended ≤ 10MB per image, formats png / jpg / webp. Overly large images may hit gateway limits. Each image in multi-image fusion must meet this limit.

How long are the returned image URLs valid? Do I need to download them?

The default url field is an R2 CDN link that expires in about 1 day (24 hours) — requests after that will 404.

Strongly recommended: download and persist generated images to your own object storage (S3 / OSS / R2), CDN, or database shortly after generation.

Does it support streaming?

No. This model returns the image in one shot; streaming is not supported. If latency matters, show a “generating…” progress indicator on the client side and configure a 300s timeout (conservative).

Can I use the official OpenAI SDK?

Yes. Point base_url to https://api.apiyi.com/v1 and set api_key to your API易 token. client.images.generate(model="gpt-image-2-vip", size="2048x1360", prompt=...) works directly.

When should I switch to the official gpt-image-2?

When you need a quality knob (low/medium/high), mask-based local repaint, or strict OpenAI-API field parity — use gpt-image-2. See the Official vs Reverse comparison.

Related Documentation

• GPT-Image-2-All Overview - Sister model at the same price with faster output, ideal when you don’t need to lock size
• ⚖️ Official vs Reverse Comparison - Side-by-side selection guide vs the official gpt-image-2 (covers -all / -vip)
• Chat Playground - Chat Completions style, one endpoint for text-to-image and editing
• Text-to-Image Playground - /v1/images/generations compatible endpoint, pass size to lock dimensions
• Image Editing Playground - /v1/images/edits multi-image fusion and editing
• GPT-Image-2 Official - For quality parameter / mask-based repaint / strict OpenAI-API field parity
• GPT-Image Series Overview - Official GPT-Image comparison
• API Manual - General calling conventions

gpt-image-2-vip is a reverse-engineered channel (Codex line). Behavior is aligned but pricing/capabilities may not fully match the official version. For full official-API parity, use gpt-image-2.