> ## Documentation Index
> Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt
> Use this file to discover all available pages before exploring further.
# GPT-image-2 Official vs Reverse
> Compare gpt-image-2 (official) vs the reverse-engineered sister models gpt-image-2-all and gpt-image-2-vip: channel nature, pricing, endpoints, upload/output formats, speed vs quality positioning, prompt adherence — pick the right one.
## TL;DR
| If you need | Pick |
| ---------------------------------------------------------------------------------------- | ---------------------------------------------------- |
| **`quality` knob / mask inpainting / locked sizes, 4K / strict OpenAI-API field parity** | `gpt-image-2` (Official, token-metered) |
| **Predictable flat \$0.03/image + fast output (speed is the advantage)** | `gpt-image-2-all` (Reverse, ChatGPT-web line, \~90s) |
| **Predictable flat \$0.03/image + sometimes higher quality (not in a hurry)** | `gpt-image-2-vip` (Reverse, Codex line, \~120–200s) |
All three models are **built on OpenAI's gpt-image-2 underneath**. The differences are in channel nature (official direct vs reverse-engineered), pricing model, and parameter granularity.
**The two reverse siblings (-all / -vip)**: This page's "Reverse" column covers **both** `gpt-image-2-all` and `gpt-image-2-vip` — they share **identical call format** and the same \$0.03/image flat price. The difference now is **speed vs quality**:
* `gpt-image-2-all`: ChatGPT web line, **\~90s** generation — **speed is the advantage**
* `gpt-image-2-vip`: Codex line, **\~120–200s** generation — slower, but **sometimes higher quality**
* Both: no `quality`, no `n`, no mask inpainting
⚠️ **`-vip`'s `size` parameter is currently broken** (since 2026-06-23, due to a Codex generation-rule change; output is fixed to adaptive 1K, no recovery ETA) — for locked sizes / 4K, use the official `gpt-image-2`. Same for `quality` tiers or mask inpainting.
**About current speeds**: `-all` / `-vip` generation is **slower than at launch** due to **OpenAI upstream compute fluctuations** — this affects all reverse-channel users, not just APIYI; our account pool and ops are healthy. Set client timeouts to 300s+ and leave more headroom for complex prompts.
## Full Comparison Table
| Dimension | **gpt-image-2-all / -vip** (Reverse, cost-effective) | **gpt-image-2** (Official) |
| -------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- |
| **Model name** | `gpt-image-2-all` (fastest) / `gpt-image-2-vip` (quality-first, when not in a hurry) | `gpt-image-2` |
| **Channel nature** | `-all`: reverse-engineered ChatGPT web line
`-vip`: reverse-engineered Codex line | Official direct (OpenAI Images API) |
| **Pricing** | **Per-call**: flat \$0.03/call (both models, same price) | **Token-metered**: matches official; \~**85% of list price** after APIYI deposit bonuses |
| **Typical cost/image** | \$0.03 (regardless of size / quality / model) | Measured **\$0.03 – \$0.2** (correlates with prompt length, size, quality) |
| **Token group** | Default | Default |
| **Token type** | **Per-call** or **Token-priority** both work | **Token-priority only** (this model is token-billed; per-call tokens will be rejected) |
| **Recommended endpoint** | **`/v1/images/generations` + `/v1/images/edits`** (more stable, more upstream supply, and same code as official — just swap the `model` name to switch during risk-control turbulence) | `/v1/images/generations` + `/v1/images/edits` |
| **Upload format** | multipart file (edits endpoint) | multipart file (edit endpoint) |
| **Output format** | `b64_json` (default, **raw base64 with no prefix**, verified 2026-07; earlier versions included the prefix) or `url` (R2 CDN) | `b64_json` (**raw base64, no prefix**) |
| **Reference image count** | Multiple | **Max 16** (`image[]`) |
| **Mask inpainting** | ❌ Not supported | ✅ Supported (alpha channel required) |
| **Prompt adherence** | Good | **Excellent** |
| **Generation speed** | `-all`: \~**90 seconds** (speed is the advantage)
`-vip`: \~**120–200 seconds** (slower, but sometimes higher quality)
📌 Currently slower than at launch — OpenAI upstream compute, not an APIYI-side issue | \~**100-120 seconds**, complex + 4K can reach 3-5 minutes |
| **Quality tendency** | `-all`: good
`-vip`: **sometimes higher** (Codex line, occasionally better detail) | Stable, and `quality=high` maxes it out |
| **`size` parameter** | `-all`: ❌ Not accepted (describe in prompt)
`-vip`: ⚠️ **Currently broken** (fixed to adaptive 1K since 2026-06-23, no recovery ETA) | ✅ Any valid custom size |
| **4K support** | `-all`: ❌
`-vip`: ⚠️ Unavailable while `size` is broken (fixed 1K) | ✅ Including `3840×2160` |
| **Common output sizes** | `-all`: 16:9 → 1672×941, 9:16 → 941×1672, 1:1 → 1254×1254 (adaptive)
`-vip`: adaptive 1K (while `size` is broken; the former [30-size table](/en/api-capabilities/gpt-image-2-vip/overview#supported-sizes-full-30-size-table) is not in effect) | 8 presets + any valid custom size |
| **`quality` parameter** | ❌ Both reverse models reject it (do not pass) | ✅ `low` / `medium` / `high` / `auto` |
| **`n` parameter** | ❌ Both reverse models reject it (1 image per call) | ✅ Supported |
| **Transparent background** | — | ❌ Not supported (`background: transparent` errors) |
| **Chinese prompts** | ✅ Native | ✅ Native |
| **Text rendering** | High fidelity | High fidelity (strongest at `high` tier) |
| **API docs** | [GPT-Image-2-All Overview](/en/api-capabilities/gpt-image-2-all/overview) / [GPT-Image-2-VIP Overview](/en/api-capabilities/gpt-image-2-vip/overview) | [GPT-Image-2 Overview](/en/api-capabilities/gpt-image-2/overview) |
🔑 **Create or manage API tokens**: [https://api.apiyi.com/token](https://api.apiyi.com/token)\
When creating a token in the console, choose a group (`Default` is fine) and a token type (**Per-call** / **Token-priority**). **Calling `gpt-image-2` (official) requires a "Token-priority" token** — per-call tokens will be rejected due to billing-mode mismatch.
## When to Pick Each
### Pick `gpt-image-2-all` (Reverse) when
Stable \$0.03/image with no size/quality tier. **Ideal for batch production with hard cost ceilings** (infographics, marketing assets, e-commerce thumbnails).
\~90s generation — **slightly faster** than both `-vip` and the official version. **Better real-time UX.**
Standard Images API format — **same code** as `-vip` and the official-relay `gpt-image-2`; switch or fall back by changing the `model` name.
Native Chinese prompt support, excellent text rendering for signage / posters / infographics — **great for Chinese-audience content production**.
### Pick `gpt-image-2-vip` (Reverse, quality-first) when
The Codex line's detail rendering is **occasionally better than `-all`** — for showcase images where you're not in a hurry and want a bit more quality at the same reverse-channel flat price.
\~**120–200s** generation, slower than `-all` — pick it when you can accept a longer wait for a higher ceiling.
Identical request structure to `-all` — **one codebase switches between both models** by swapping the `model` name based on your speed / quality preference.
Same flat \$0.03/image as `-all` — batch production costs stay capped.
**`-vip`'s former "locked sizes / 4K" selling point does not currently hold**: the `size` parameter has been broken since 2026-06-23 (a Codex generation-rule change; output is fixed to adaptive 1K, no recovery ETA). For e-commerce hero shots, poster templates, 4K wallpapers and any **locked-size / 4K needs, use the official `gpt-image-2`**.
### Pick `gpt-image-2` (Official) when
`quality` supports low/medium/high/auto. **Use `low` for drafts to save cost; `high` for print-grade finals** — official-only; both reverse models reject it.
Alpha-channel mask supported — **precisely modify a region while preserving the rest**. Both reverse models do not support this.
`size` accepts any valid resolution (including 4K). While the reverse channel's `size` is broken, **every exact-dimension or 4K workload goes official**.
Goes through the official Images API — fields and behavior identical to OpenAI official. **Existing OpenAI-SDK-based code / systems migrate with zero changes** and stay stable long-term.
## Key Differences in Detail
### 1. b64\_json format gotcha (migration trap!)
As verified in July 2026, both models now return **raw base64 (no `data:` prefix)** — but `gpt-image-2-all` used to include the prefix, so the safest shared code checks for it first:
```python theme={null}
# Universal pattern: detect the prefix before processing — works for both models
b64 = resp["data"][0]["b64_json"]
if b64.startswith("data:"): # handles historical prefixed responses
b64 = b64.split(",", 1)[1]
with open("out.png", "wb") as f:
f.write(base64.b64decode(b64)) # ✅ write file
img_tag = f'
' # ✅ browser render
```
When switching between the two, **the `b64_json` handling code must change**, or you'll get a corrupted data URL or a decode failure.
### 2. Resolution control
**gpt-image-2-all** (in the prompt):
```
"Landscape 16:9 cinematic, old lighthouse at sunset" → ~1672×941
"Portrait 9:16 phone wallpaper, cyberpunk city" → ~941×1672
"1024×1024 square logo, minimalist cat line art" → ~1254×1254
```
**gpt-image-2-vip** (`size` currently broken, since 2026-06-23):
It used to accept 30 explicit sizes (including 4K), but after a Codex generation-rule change **the `size` parameter is broken and output is fixed to adaptive 1K**, with no recovery ETA. For now, describe the composition in the prompt just like `-all`; **for exact dimensions / 4K, use the official `gpt-image-2`**.
**gpt-image-2** (`size` parameter strict + `quality` tiers):
```python theme={null}
client.images.generate(
model="gpt-image-2",
prompt="...",
size="2048x1152", # ✅ output exactly this
quality="high" # official-only
)
```
### 3. Upload / output format differences
| Operation | gpt-image-2-all | gpt-image-2 |
| ---------------------- | ------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------------- |
| **Upload reference** | multipart `image` file field (edits endpoint) | multipart `image[]` file field |
| **Download output** | Default `b64_json` (raw base64, verified 2026-07); explicit `response_format: "url"` returns an R2 CDN link (**24h validity**) | `b64_json` (**raw base64**, requires decode) |
| **Multi-image fusion** | Repeat the `image` field on the edits endpoint | `image[]` array, **max 16** |
### 4. Cost ballpark
| Scenario | gpt-image-2-all / -vip | gpt-image-2 |
| ------------------------- | --------------------------------------------------------- | ----------------------------------------------------- |
| 1024×1024 draft | \$0.03 | \~\$0.006 (low) |
| 1024×1024 medium quality | \$0.03 | \~\$0.053 (medium) |
| 1024×1024 high quality | \$0.03 | \~\$0.211 (high) |
| 2048×1152 high quality | \$0.03 | \~\$0.20+ (token-metered) |
| 3840×2160 4K high quality | — (no reverse model can target 4K while `size` is broken) | Token-metered, **significantly higher than 1K** |
| Edit / multi-image fusion | \$0.03 | Input tokens rise sharply, single call can hit \$0.1+ |
**Bottom line**: For batch / low-quality workloads, the reverse channel isn't always cheaper (1K low is actually less expensive on the official tier). The **mid-to-high quality range** is the reverse channel's \$0.03 sweet spot. Pick official (token-metered) when you need **`quality` tiers / mask inpainting / locked sizes, 4K / strict OpenAI-API field parity**.
## Client Settings
| Setting | gpt-image-2-all / -vip | gpt-image-2 |
| -------------------------- | ------------------------------------------------------------------------- | ------------------------------------------------------------ |
| **Timeout (conservative)** | `-all`: **300s** (typical \~90s)
`-vip`: **300s** (typical 120–200s) | **360s** (4K high quality realistically reaches 3-5 minutes) |
| **Retry strategy** | Exponential backoff on 5xx / timeout, max 2 retries | Same |
| **Concurrency** | 1 image per call — issue parallel requests for multiple | 1 image per call — issue parallel requests for multiple |
| **Request ID** | `request-id` response header | `x-request-id` response header |
**Common to all three models: for image edit / multi-image fusion, 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 — compressing measurably improves success rate and latency. **Output resolution is independent of input size** — quality is set on the output side (`size` + `quality` for official; prompt phrasing for `-all` and for `-vip` while its `size` is broken), not by input file size.
## FAQ
**Yes, strongly recommended.** For all three models, 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.
**Don't worry about compression hurting quality** — output resolution is independent of input size. The "output-side" controls differ across the three:
* `gpt-image-2-all`: controlled by prompt composition phrasing (see the verified phrasing table on the [-all overview page](/en/api-capabilities/gpt-image-2-all/overview)) — `4K` / `8K` in the prompt does not count
* `gpt-image-2-vip`: the `size` field is **currently broken** (fixed to adaptive 1K) — put composition intent in the prompt too
* `gpt-image-2`: controlled by `size` + `quality` (any valid size)
Bottom line: shrinking inputs only speeds things up — quality is set by output-side configuration, not input file size.
Yes. All three run on the Default channel — the same API Key calls them with no extra config. Note: calling `gpt-image-2` (official) requires a "Token-priority" token; `-all` / `-vip` accept either token type.
**Use the OpenAI Images API** (`/v1/images/generations` for text-to-image + `/v1/images/edits` for editing), for two reasons:
1. **More stable**: upstream resource supply for the Images API channel is more plentiful, so call success rates are higher
2. **Compatible with the official relay for easy switching**: the call method and parameter format are fully compatible with the official-relay `gpt-image-2` — if the reverse channel hits risk-control turbulence, **just swap the `model` name to switch to the official relay** with zero code changes
There is also a chat-based endpoint (`/v1/chat/completions`, **no longer recommended**), only useful for multi-turn iterative editing or passing online image URLs directly. Note that when the image intent is ambiguous, it may return plain text instead of an image (prepend a fixed prefix like "Generate an image:" to reinforce it). For full parameters, see the [-all chat-based API reference](/en/api-capabilities/gpt-image-2-all/chat-completions) / [-vip chat-based API reference](/en/api-capabilities/gpt-image-2-vip/chat-completions).
Both are reverse-engineered channels at the same flat price (\$0.03/image), with **identical call format** (`-vip`'s `size` is currently broken, so neither takes `size`). The difference is **speed vs quality**:
* **Generation time**: `-all` \~90s — **speed is the advantage**; `-vip` \~120–200s. Currently slower than at launch due to OpenAI upstream compute fluctuations
* **Quality**: `-vip` (Codex line) detail rendering is **sometimes higher** — for showcase images when you're not in a hurry
Decision: want fast output → `-all`; quality-first, not in a hurry → `-vip`; need locked sizes or 4K → official `gpt-image-2`. See the [GPT-Image-2-VIP Overview](/en/api-capabilities/gpt-image-2-vip/overview) for details.
Use the official `gpt-image-2`. `-vip`'s `size` parameter has been broken since 2026-06-23 (fixed to adaptive 1K, no recovery ETA), so neither reverse model can precisely control output dimensions right now.
Official-only features: **any valid size (incl. 4K)**, **`quality` tiers** (low/medium/high/auto), **mask inpainting** (alpha-channel mask), **strict OpenAI-API field parity** (zero-change migration for existing OpenAI-SDK code). Token-metered billing.
* **Stick with the OpenAI SDK / must match OpenAI official, or need locked sizes / 4K**: pick `gpt-image-2` (official). Drop `input_fidelity`, avoid `background: transparent`, leave the rest unchanged.
* **Cut cost, want fast output**: pick `gpt-image-2-all` (reverse, \~90s).
* **Cut cost, quality-first and not in a hurry**: pick `gpt-image-2-vip` (reverse, \~120–200s).
Yes. A common pattern: **primary `-all` or `-vip`** (predictable cost — pick by speed / quality preference), **fallback `gpt-image-2`** (switch when you need `quality` tiers, mask, or locked sizes). The reverse and official response shapes differ — normalize at the business layer.
See [Slow CDN downloads — what to do](/en/faq/cdn-download-slow)
## Related Docs
* [GPT-Image-2 Overview](/en/api-capabilities/gpt-image-2/overview) - Full official integration docs
* [GPT-Image-2-All Overview](/en/api-capabilities/gpt-image-2-all/overview) - Reverse ChatGPT-web line (fastest output) full integration docs
* [GPT-Image-2-VIP Overview](/en/api-capabilities/gpt-image-2-vip/overview) - Reverse Codex line (sometimes higher quality; `size` currently broken) full integration docs
* [Deep dive: gpt-image-2 launch](/en/news/gpt-image-2-launch) - Official version launch
* [Deep dive: gpt-image-2-all launch](/en/news/gpt-image-2-all-launch) - Reverse-engineered version launch
* [Community: Luck GPT-Image 2 ComfyUI Nodes](/en/scenarios/ecosystem/luckgpt2-comfyui) - Multi-model ComfyUI node pack
* [Community: APIYI GPT-Image 2 Skills](/en/scenarios/ecosystem/apiyi-gpt-image-skills) - Multi-model AI Agent Skill pack
* [Deposit promotions](/en/faq/recharge-promotions) - Recharge bonus policy