Seedream 5.0 Lite在ModelGo：四组真实API测试，费用降低30.7%

四组不同的生产需求、两个API平台，以及实际费用记录。本组测试中，ModelGo合计费用为$0.097059，OpenRouter为$0.140000；在实际交付规格一致的条件下，ModelGo费用低30.67%。

Seedream 5.0 Lite是什么？

字节跳动Seedream 5.0 Lite将图像生成与参考图编辑整合在同一工作流中。适合商品目录排版、材质细节展示和成套产品变体设计。

测试方法与实际费用

本次测试包含四组独立需求：两组纯文字生成、两组参考图生成；在双平台可访问时，使用相同Prompt、参考文件和请求设置。ModelGo费用来自逐条匹配的已结算账单，OpenRouter费用来自已完成响应中的用量费用；表格展示模型输出费用，保留六位小数，不包含单独生成参考图的费用。

费用口径：本文比较模型调用费用，不是为账户充值所需的实际支出；不包含充值或支付手续费，包括另一篇价格对比文章计入的5.5%充值费。本文费用差额及比例仅按调用费用计算。

四个真实生产案例

为准确复现真实调用，本节保留当时提交的英文Prompt原文；中文说明用于解释生产目标。

三套胶囊系列穿搭

应用行业：时尚/商品企划。

制作同一胶囊系列的三套穿搭展示，检验系列配色、服饰材质、人物布局和商品企划表达；完整制作要求见英文原文。

实际提交的英文Prompt

Create a premium fashion merchandising board for fictional label “MARGIN STUDIO”. Show the same synthetic adult model three times in a consistent neutral pose, wearing three coordinated outfits: a sand linen suit, a charcoal knit dress and an olive technical coat over trousers. Keep the model’s identity, height, skin tone and proportions consistent, while making the garments clearly distinct. Use a warm-gray seamless studio backdrop, soft side light, realistic fabric texture and equal spacing between the three full-body views. Include small editorial labels “01 Linen”, “02 Knit”, “03 Shell” beneath the respective looks. No cropped feet, merging silhouettes, real designer logos or unrelated accessories. Deliver one finished image, not a collage of alternative answers. Respect the specified aspect ratio and keep every requested object and text region distinct.

ModelGo费用低30.67%。

家庭套餐菜单板

应用行业：快餐/门店展示。

设计可用于快餐门店的家庭套餐菜单板，观察菜品与文案信息分区、商品表现及展示可读性；精确文案按实际Prompt保留。

实际提交的英文Prompt

Design a polished digital menu board for fictional neighborhood kitchen “SUNDAY TABLE”. Use an overhead hero arrangement of roast vegetables, a baked pie, a green salad and a fruit crumble, each in a different realistic serving dish on warm oak. On the left third create a tidy typographic column with exact copy “THE FAMILY TABLE”, “A meal to share”, and the three entries “Garden pie”, “Roast vegetables”, “Apple crumble”. Keep the food imagery abundant but believable, with consistent shadows and no utensils penetrating food. Use cream and tomato-red graphic accents, generous margins and legible text. Do not invent dietary certifications, discounts or nutritional statistics. Deliver one finished image, not a collage of alternative answers. Respect the specified aspect ratio and keep every requested object and text region distinct.

ModelGo费用低30.67%。

移动电源配件详情页

应用行业：消费电子/电商。

使用参考移动电源制作配件详情页，检查产品身份、功能部位和信息标注之间的一致性。

这张原创参考图用于ModelGo；对OpenRouter可访问的模型，对照请求也使用同一文件。

实际提交的英文Prompt

Use the supplied graphite rectangular power bank as the exact product reference. Preserve its rounded corners, two visible ports, small circular indicator and matte finish. Build a square ecommerce detail-page composition with the power bank at a three-quarter angle in the center, a neatly coiled charging cable at lower left and two enlarged circular detail windows showing the ports and indicator. Use thin non-crossing leader lines and the labels “Dual connection”, “Status indicator” and “Pocket format”. Keep all ports physically plausible and consistent between views. Work on a warm off-white background with soft studio lighting. Do not invent capacity, charging wattage, compatibility badges or regulatory marks. Deliver one finished image, not a collage of alternative answers. Respect the specified aspect ratio and keep every requested object and text region distinct.

ModelGo费用低30.67%。

模块化玩具成套包装

应用行业：玩具/包装设计。

依据原创玩具参考图设计成套包装，观察系列视觉统一、结构保留和包装展示布局。

这张原创参考图用于ModelGo；对OpenRouter可访问的模型，对照请求也使用同一文件。

实际提交的英文Prompt

Use the supplied original wooden modular creature toy as the exact reference, retaining its round head, three detachable body segments and painted teal feet. Design a presentation of three coordinated retail cartons for fictional brand “TINY FORMS”, with colorways Coral, Moss and Sky. Each carton should show a large product photo window, a small diagram of the same three-segment construction and its colorway name. Arrange the cartons at consistent three-quarter angles on a pale-yellow studio surface with one assembled toy in front. Keep toy geometry consistent across packaging artwork and the physical toy. Omit age ratings, safety certifications, licensed characters and unsupported educational claims. Deliver one finished image, not a collage of alternative answers. Respect the specified aspect ratio and keep every requested object and text region distinct.

ModelGo费用低30.67%。

如何计算费用差额

对于实际交付规格一致的输出，费用降低比例的计算方式为(OpenRouter 实际费用 − ModelGo 已结算费用) / OpenRouter 实际费用 × 100。判断两次输出是否可比时，应检查实际返回的尺寸、时长和音轨。这四组需求的平均单次费用差为$0.010735；如果按同样工作负载重复100或1,000次，费用差约为$1.073529或$10.735294。这只是按本次实测推算的用量示例，不是新增实际账单。

开始使用ModelGo API

先安装requests并设置环境变量MODELGO_API_KEY。以下代码使用首个真实案例的端点、模型ID和参数，已在代码中直接写入实际Prompt。

import os, requests

prompt = 'Create a premium fashion merchandising board for fictional label “MARGIN STUDIO”. Show the same synthetic adult model three times in a consistent neutral pose, wearing three coordinated outfits: a sand linen suit, a charcoal knit dress and an olive technical coat over trousers. Keep the model’s identity, height, skin tone and proportions consistent, while making the garments clearly distinct. Use a warm-gray seamless studio backdrop, soft side light, realistic fabric texture and equal spacing between the three full-body views. Include small editorial labels “01 Linen”, “02 Knit”, “03 Shell” beneath the respective looks. No cropped feet, merging silhouettes, real designer logos or unrelated accessories. Deliver one finished image, not a collage of alternative answers. Respect the specified aspect ratio and keep every requested object and text region distinct.'
headers = {"Authorization": "Bearer " + os.environ["MODELGO_API_KEY"]}
payload = {'model': 'doubao-seedream-5.0-lite',
 'prompt': prompt,
 'n': 1,
 'size': '2560x1440',
 'response_format': 'b64_json'}
response = requests.post('https://api.modelgo.ai/v1/images/generations', headers=headers, json=payload, timeout=420)
response.raise_for_status()
result = response.json()

本示例使用b64_json返回格式，可解码响应中data[0].b64_json的图像数据。

本次Seedream参考图请求通过已测试的生成端点，在image字段传入参考图的data URL。

ModelGo与OpenRouter的模型ID和媒体请求格式存在差异。可以复用创作需求，接入时请根据各平台文档调整端点、参考图结构、任务轮询和下载处理。

ModelGo API文档

在ModelGo使用Seedream 5.0 Lite

使用本页介绍的Seedream 5.0 Lite模型，请前往ModelGo的Models页面。

前往ModelGo使用Seedream 5.0 Lite

参考资料

以下资料用于核对模型能力、接口与用量记录；案例需求由本次测试独立编写。

• 模型官方文档或系列展示：用于了解能力与创意方向；相关系列资料不代表其他变体的规格保证。
• OpenRouter API参考：用于核对请求约定及平台差异。
• OpenRouter用量与费用记录：用于理解用量信息；本文费用来自实际测试记录。
• OpenRouter模型目录API：在新接入前核对模型ID与目录属性。
• OpenRouter图像生成指南：用于了解媒体请求结构和工作流。