OpenAIImage model检查中
GPT Image 2.6 API
gpt-image-2-6-sunburst-text-to-image-stable-tokenGenerate or edit images with GPT Image 2.6: choose Flare for fast iteration or Sunburst for detailed visual work.
- 文生图
- 图生图
- 动态 GIF
- SVG 输出选项
- Up to 16 input images
- 标注引导编辑
- 质量选项
在 Playground 中运行 GPT Image 2.6
The form is the request body: every field shows its key. Sign in to run it; a run is billed like an API task.
参数
modelGPT Image 2.6Sunburst text to image tokengpt-image-2-6-sunburst-text-to-image-stable-token
高级参数 (4)
input.sizeOutput size. Use auto, or a supported size such as 1024x1024, 1536x1024, or 1024x1536. A custom size is WIDTHxHEIGHT: both edges must be multiples of 16 and the aspect ratio must be between 1:3 and 3:1. Neither edge may exceed 3840, and total pixels must be between 655360 and 8294400. Sizes above 2560x1440 are experimental. Omit to use auto. ≤ 32 chars · 4/32
input.frameCountGIF frame count, 2 to 16. Set only when outputFormat is gif. Upstream defaults to 9 when omitted. 2–16
input.frameDurationMsGIF frame duration in milliseconds, 20 to 2000, rounded to 10ms. Set only for gif. Upstream defaults to 200. 20–2000
input.outputCompressionOutput compression, 0 to 100. Applies to jpeg and webp only. Omitted when not sent. 0–100
任务接收时预留额度,仅成功后扣费。失败任务释放预留额度;结果未确认时会保留额度,直到核实。
请求预览
{"model": "gpt-image-2-6-sunburst-text-to-image-stable-token","input": {"size": "auto","background": "auto","quality": "auto","outputFormat": "png"}}网页 Playground 使用你的登录会话发送此请求体。在代码中请使用 API Key 发送相同内容。
结果
运行请求后,结果会显示在这里,并附上状态和费用。
GPT Image 2.6 价格
各可用版本和规格的价格以美元显示,计费单位随费率注明;只有找到可匹配的官方标价时才显示参考价格。
- Charged only on success仅对成功任务扣费。失败任务释放预留额度;结果尚未确认的任务会保留额度,直到结果核实。
- Safe retries使用相同幂等密钥重试不会重复扣费。
- Price changes发送 expectedPriceMicros:若展示后价格发生变化,任务会以 PRICE_CHANGED 被拒绝。
对比 GPT Image 2.6 各版本
每个变体都有独立的模型 ID 和输入参数。请先选择任务和版本;变体存在差异时会显示额外字段。
Sunburst text to image token默认值
gpt-image-2-6-sunburst-text-to-image-stable-token文生图; 文本
gpt-image-2-6-sunburst-image-to-image-stable-token图生图; 文本, 图像
gpt-image-2-6-flare-text-to-image-stable-token文生图; 文本
gpt-image-2-6-flare-image-to-image-stable-token图生图; 文本, 图像
| 版本 | 任务类型 | 输入 |
|---|---|---|
Sunburst text to image token默认值gpt-image-2-6-sunburst-text-to-image-stable-token | 文生图 | 文本 |
Sunburst image to image tokengpt-image-2-6-sunburst-image-to-image-stable-token | 图生图 | 文本, 图像 |
Flare text to image tokengpt-image-2-6-flare-text-to-image-stable-token | 文生图 | 文本 |
Flare image to image tokengpt-image-2-6-flare-image-to-image-stable-token | 图生图 | 文本, 图像 |
GPT Image 2.6 能做什么
OpenAI positions Flare for fast, interactive image generation and editing, while Sunburst is intended for detailed compositions and image editing. Both accept text and image inputs. Select the variant and task based on your own quality and latency needs; results and timing vary by request.
The OpenAI image guide describes adjustable generation complexity, markup-guided edits, animated GIF generation and SVG output. Complexity and rendering quality are separate settings. OpenAI documents SVG for image generation, but not image editing.
Animated GIF is a distinct output option: use it to explore short motion concepts from a text prompt or an edit reference. OpenAI documents GIF generation and frame controls in its image guide; the Hermes schemas list frameCount and frameDurationMs. The manufacturer documents 2–16 frames and 20–2,000 ms per frame, subject to a total pixel budget; its GIF output does not support streaming or custom compression. Its API reference also describes preview access and mask restrictions. Check availability before running and validate the delivered result for your workflow.
The Hermes request schema has four variants: Flare and Sunburst, each with text-to-image and image-to-image modes. Edit requests list one to 16 reference images in imageUrls, plus optional maskUrl and markupUrl. Text-to-image requests list PNG, JPEG, WebP, GIF and SVG; edit requests list PNG, JPEG, WebP and GIF. These are request options, not a guarantee that every format is currently available for delivery.
Check the live availability and pricing on this page before sending a task. Input text, reference images, output format and quality can affect upstream usage; the price and billing rules for a variant are shown when configured.
能力
动态 GIF
请求参数包含动态 GIF 输出以及帧数和帧时长设置。运行前请先确认当前可用状态。
outputFormatframeCountframeDurationMsEvery variant
SVG 输出选项
此变体的请求参数包含 SVG 输出。运行前请先确认当前可用状态和结果交付方式。
outputFormat版本: Sunburst text to image token, Flare text to image token
标注引导编辑
编辑参数支持使用标注图指出原图中需要修改的位置。
markupUrlmaskUrl版本: Sunburst image to image token, Flare image to image token
质量选项
选择生成结果的质量;可选项因变体而异。
auto · low · medium · high · xhigh · max
qualityEvery variant
格式选项
请求参数列出了这些格式;具体可用性和结果交付取决于所选变体。
JPEG · PNG · WEBP · GIF
outputFormatEvery variant
Up to 16 input images
Send up to 16 images in one request as HTTPS URLs or uploaded media references.
imageUrls版本: Sunburst image to image token, Flare image to image token
使用场景
Explore ideas quickly
Try Flare for interactive iteration on a written brief, then compare settings against the result you need.
Refine detailed compositions
Try Sunburst for work where the composition and editing details matter.
Edit from references
Supply up to 16 images and describe what to keep or change. The edit schema includes optional masks and markup.
Prepare motion or vector concepts
The manufacturer documents GIF generation and SVG output for supported requests. Check the selected Hermes variant's current availability before using these formats.
用代码调用 GPT Image 2.6
Every image, video and music model uses the same asynchronous job API: submit a task, then poll for the result.
- 1
创建 API Key
In the dashboard, with the jobs:create and jobs:read scopes. API 密钥 - 2
提交任务
POST /api/v1/jobswith the model ID and an input object. The response is 202 with a task ID. - 3
Poll for the result
GET /api/v1/jobs/{id}until the status is succeeded, then download the output URLs.
最小请求
Required fields only, generated from the schema of Sunburst text to image token. Set HERMES_API_KEY in your environment.
- 模型 ID
gpt-image-2-6-sunburst-text-to-image-stable-token- 端点
https://api.hermes-ai.net/api/v1/jobs
# 1. Submit the task (asynchronous; returns 202 with a task id)
curl -X POST https://api.hermes-ai.net/api/v1/jobs \
-H "Authorization: Bearer $HERMES_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{
"model": "gpt-image-2-6-sunburst-text-to-image-stable-token",
"input": {
"prompt": "A lighthouse on a rocky coast at sunrise, soft light"
}
}'
# 2. Poll until status is succeeded, failed or reconciliation_required
curl https://api.hermes-ai.net/api/v1/jobs/TASK_ID \
-H "Authorization: Bearer $HERMES_API_KEY"输入参数
Fields of Sunburst text to image token (gpt-image-2-6-sunburst-text-to-image-stable-token). 未知字段会被拒绝并返回 INVALID_MODEL_INPUT。
| 字段 | 类型 | 必填 | 默认值 | 可选值与约束 |
|---|---|---|---|---|
prompt | string | 必填 | — | Image generation prompt (2-32000 characters). 2–32000 chars |
size | string | 可选 | "auto" | Output size. Use auto, or a supported size such as 1024x1024, 1536x1024, or 1024x1536. A custom size is WIDTHxHEIGHT: both edges must be multiples of 16 and the aspect ratio must be between 1:3 and 3:1. Neither edge may exceed 3840, and total pixels must be between 655360 and 8294400. Sizes above 2560x1440 are experimental. Omit to use auto. ≤ 32 chars |
background | string | 可选 | "auto" | Output background mode. 可选值: |
quality | string | 可选 | "auto" | Generation quality; higher levels increase latency and token usage. 可选值: |
complexity | string | 可选 | — | Planning effort before generation. low, medium, high. Omitted when not sent. 可选值: |
moderation | string | 可选 | — | Moderation level. auto or low. Omitted when not sent; upstream defaults to auto. 可选值: |
svgModel | string | 可选 | — | Text model that writes the SVG. Set only when outputFormat is svg. Omitted otherwise. 可选值: |
svgReasoningEffort | string | 可选 | — | Reasoning effort for the SVG author. Applied only when outputFormat is svg and svgModel is set. Omitted when not sent. 可选值: |
frameCount | integer | 可选 | — | GIF frame count, 2 to 16. Set only when outputFormat is gif. Upstream defaults to 9 when omitted. 2–16 |
frameDurationMs | integer | 可选 | — | GIF frame duration in milliseconds, 20 to 2000, rounded to 10ms. Set only for gif. Upstream defaults to 200. 20–2000 |
outputCompression | integer | 可选 | — | Output compression, 0 to 100. Applies to jpeg and webp only. Omitted when not sent. 0–100 |
outputFormat | string | 可选 | "png" | Output image format. 可选值: |
Headers, response, task statuses and errors
| 请求头 | 值 | 说明 |
|---|---|---|
Authorization | Bearer <API key> | 也可以通过 x-api-key 发送密钥。 |
Content-Type | application/json | |
Idempotency-Key | <unique string ≤128> | 必填。仅在重试完全相同的请求时复用。 |
| 字段 | 类型 | 说明 |
|---|---|---|
id | string | Hermes AI 任务 ID,用于轮询。 |
model | string | 模型 slug。 |
status | string | 参见任务状态。 |
errorCode | string | null | 任务失败或需要核对时设置。 |
priceMicros | integer | 接收时锁定的报价,美元微单位。 |
currency | "USD" | |
createdAt / updatedAt | string | ISO 8601 时间戳。 |
assets[] | array | { id, type, url?, text? }。url 为同源 /api/v1/results/<uuid> 链接。 |
replayed | boolean | 仅提交时返回。幂等重试命中时为 true。 |
| 状态 | 终态 | 含义 |
|---|---|---|
queued | 否 | 已接收,等待执行。 |
submitting | 否 | 正在提交给模型提供方。 |
running | 否 | 模型生成中。 |
succeeded | 是 | 已完成,结果位于 assets。 |
failed | 是 | 未产生结果,预留额度已释放。 |
reconciliation_required | 是 | 结果正在核实。请保留任务 ID,不要自动重新提交。 |
| HTTP | 代码 | 含义与处理方式 |
|---|---|---|
401 | AUTH_REQUIRED | 缺少或无效的 API Key 或会话。 |
403 | INSUFFICIENT_SCOPE | API Key 缺少此端点所需的权限。 |
403 | PERMISSION_DENIED | 此凭据无权执行该操作。 |
429 | API_KEY_RATE_LIMITED | 已达到 API Key 的频率限制,请遵循 Retry-After。 |
429 | API_KEY_USAGE_EXCEEDED | API Key 已达到使用上限,请创建新密钥。 |
400 | INVALID_REQUEST | 请求体不是 JSON 对象,或缺少必需值。 |
400 | INVALID_IDEMPOTENCY_KEY | Idempotency-Key 必须是 1–128 个可打印 ASCII 字符。 |
400 | INVALID_MODEL_INPUT | input 与模型 schema 不匹配(未知字段、缺少必填字段或值无效)。 |
400 / 413 | INPUT_TOO_LARGE | 请求体超过 256 KB。请先上传媒体并发送媒体引用。 |
404 | REQUEST_NOT_FOUND | No chat request with this ID belongs to your account. |
404 | MODEL_NOT_FOUND | 没有此 slug 对应的模型。 |
404 | JOB_NOT_FOUND | 你的账户中没有此 ID 的任务。 |
409 | IDEMPOTENCY_CONFLICT | 此 Idempotency-Key 已用于不同的输入。新请求请使用新密钥。 |
409 | PRICE_CHANGED | 价格与 expectedPriceMicros 不一致。请重新获取价格并确认。 |
409 | MODEL_PRICE_UNAVAILABLE | 此模型尚未配置价格,暂不可运行。 |
409 | PAYMENT_REVIEW_REQUIRED | 退款后余额不足。请联系支持或补足余额后再提交任务。 |
409 | INSUFFICIENT_CREDITS | 可用额度不足以支付此任务。 |
429 | RATE_LIMITED | 每分钟提交次数过多,请稍后重试。 |
429 | EXPORT_RATE_LIMITED | 本分钟内导出 CSV 次数过多,请按 Retry-After 等待后重试。 |
429 | CONCURRENCY_LIMITED | 未完成任务过多,请等待任务完成。 |
503 | EXECUTION_UNAVAILABLE | 此模型当前未配置执行能力。 |
429 | UPSTREAM_RATE_LIMITED | The model provider is rate limiting chat requests. Retry with backoff; nothing was charged. |
502 | UPSTREAM_ERROR | The model provider failed the chat request. Nothing was charged; retry later. |
503 | PLATFORM_BUDGET_EXCEEDED | 此模型暂时不可用,请稍后重试。 |
503 | EXECUTION_TEMPORARILY_UNAVAILABLE | 执行暂时不可用,请使用相同的 Idempotency-Key 重试。 |
503 | AUTH_UNAVAILABLE | 身份验证暂时不可用,请稍后重试。 |
503 | API_UNAVAILABLE | API 暂时不可用,请稍后重试。 |
GPT Image 2.6 API 常见问题
Answers built from the model's schema and the Hermes AI API contract.
Which variant should I choose?
Pick text to image for a written prompt, or image to image to supply references. OpenAI positions Flare for fast iteration and Sunburst for detailed visual work. Check the current status and price of the exact Hermes variant before running it.
Why can't I run a variant?
A listed variant may be unavailable while its pricing or execution is being configured. Its current status and the playground's run control show whether it can accept requests.
Can I provide a mask?
The image-to-image variants list optional maskUrl and markupUrl fields. The source describes a mask as a PNG with an alpha channel and the same dimensions as the source image.
What can the GPT Image 2.6 API do?
OpenAI 的 GPT Image 2.6 支持文生图 和 图生图。目录列有4 variants, each,每个变体都有自己的模型 ID 和输入参数。提交请求前请先确认当前可用状态。
What input does GPT Image 2.6 accept?
A text prompt of up to 32,000 characters, plus 图像 input (up to 16 images per request). Media fields take a public HTTPS URL or a media:<uuid> reference returned by POST /api/v1/media, so local files can be uploaded first. 未知字段会被拒绝并返回 INVALID_MODEL_INPUT。
How is GPT Image 2.6 priced and billed?
You pay per successful task from prepaid credits, with no subscription. The price depends on the variant and the specification you choose, such as resolution or duration; see the pricing table on this page or the pricing page. Credits are reserved when a task is accepted and charged only when it succeeds; a failed task returns the reservation to your balance.
Which GPT Image 2.6 variant should I choose?
For 文生图, Sunburst text to image token 和 Flare text to image token take the same input fields and differ in tier and price; for 图生图, Sunburst image to image token 和 Flare image to image token take the same input fields and differ in tier and price. Pick the task first, then compare the tiers in the variants table.
How do I call the GPT Image 2.6 API from code?
Send POST https://api.hermes-ai.net/api/v1/jobs with the variant's model ID and an input object, then poll GET /api/v1/jobs/{id} until the status is succeeded or failed. The API section has cURL, JavaScript and Python examples generated from the schema.
Can I try GPT Image 2.6 without writing code?
Yes. Sign in and use the playground on this page: it sends the same request as the API with your session and charges the same price. Code needs an API key from the dashboard.
What happens when a task fails or I hit a rate limit?
A failed task is not charged and its reserved credits are returned. Too many submissions per minute return 429 RATE_LIMITED, and too many unfinished tasks return 429 CONCURRENCY_LIMITED; retry later or wait for a task to finish.