Segment Anything Model 3.1
Meta 的分割模型，可根据简短文本提示返回图像和视频中对象的边界框和掩码。
规格
- 上下文
- —
- 最大输出
- —
- 输入价
- —
- 输出价
- —
- 发布日期
- 2026-03-27
上下文
- 官方未公布
价格
- 官方未公布
输入输出
- 输入
- 文本图像视频
- 输出
- 文本
API
- 接口类型
思考
- 官方未公布
源信息
- 状态
- 可用
- 发布日期
- 2026-03-27
- 知识截止
- —
能力
- 已确认
调用方式
1 个 Provider
metamodel = sam-3.1
responses
POSThttps://api.meta.ai/v1/v1/responses
JSON
标准格式
{
"id": "sam-3.1",
"object": "model",
"created": 1774569600,
"owned_by": "meta",
"name": "Segment Anything Model 3.1",
"api": {
"types": [
"responses"
]
},
"limits": {
"context": null,
"input": null,
"output": null
},
"modalities": {
"input": [
"text",
"image",
"video"
],
"output": [
"text"
]
},
"reasoning": {
"supported": null,
"efforts": []
},
"pricing": null,
"features": [
"streaming"
],
"info": {
"status": "active",
"release_date": "2026-03-27",
"knowledge_cutoff": null,
"description": "Meta's segmentation model that returns boxes and masks for objects in images and video from a short text prompt.",
"docs": "https://dev.meta.ai/docs/media-segmentation",
"verified_at": "2026-10-11"
}
}官方来源
4