白嫖新大陆:这两家AI平台免费到底The New World of Freebies: These Two AI Platforms Are Completely Free


AI free token/AI免费积分

 

 💰 Vol.004 · 白嫖攻略 · 2026.06.25  ⏱ 阅读约 7 分钟

 

 

   📍 本文阅读进度    开始探索  

 

   
 
 

 

   🔍 商汤🤖 Agnes🛠 实操⚠️ 避坑🎯 总结  

 

 😱 AI 工具费越来越贵了。ChatGPT 每月 160,Claude 每月 160,Midjourney 每月 200+,想做 AI 副业的人还没开始赚钱,工具费先交了大几百。

 

 💡 但上周我挖到了两个真正的宝藏平台——商汤日日新和Agnes AI。两个平台加起来六款大模型,覆盖写稿、生图、做视频三大场景,全部免费。

 

 

   ⚡ 一句话定位:这篇文章是全网最详细的「商汤+Agnes 双平台白嫖指南」——从注册到代码调用,从避坑到省钱账,看完就能上手。所有信息基于 2026 年 6 月真实测试。  

 

 🔵 Part 1:六款免费模型,一张表看清楚

 

 ⏰ 本部分阅读约 1 分钟

 

                                                                                                                                                                                                                                                                                                                                                                                                                       

平台 模型 擅长 类型 免费规则 推荐
商汤 Flash-Lite 全能打杂 文本 1500次/5h ⭐⭐⭐⭐⭐
U1 Fast 信息图/海报 生图 1500次/5h ⭐⭐⭐⭐
DeepSeek V4 深度推理 文本 150-500次/5h ⭐⭐⭐⭐
Agnes 2.0-flash 日常对话 文本 永久免费 ⭐⭐⭐⭐⭐
image-2.0 AI 生图 生图 永久免费 ⭐⭐⭐⭐⭐
video-v2.0 AI 视频 视频 永久免费 ⭐⭐⭐⭐

 

 

   一句话:六款模型,从写稿、翻译、做图、分析数据,到生成短视频,日常AI场景全包了——关键是一个子儿不用花。  

 

 🔴 Part 2:商汤日日新 —— 全能打杂王 + 深度推理

 

 ⏰ 本部分阅读约 3 分钟

 

                   

🏠 入口 token.sensenova.cn
🔗 API格式 完全兼容 OpenAI
💰 费用 三款模型各自免费 1500 次/5 小时
📋 注册 手机号收验证码,无需实名/绑卡

 

 商汤给的是三款模型,各自独立计费——一个用完不扣另一个的额度。

 

 

① Flash-Lite · 多模态全能打杂王

 

   能读文档、看图片、识别表格、分析 OCR。支持 252K 上下文,一次扔一本小说进去都能读懂。日常办公的绝对主力。  

     

💰 1500 次 / 5 小时 · 敞开用

 

 

 

② U1 Fast · 信息图/海报创作器

 

   专攻信息图、海报、数据可视化。输入数据直接出柱状图/饼图/流程图,还能生成 PPT 级别的排版。不会设计也能秒变「会做图」。  

     

💰 1500 次 / 5 小时 · 敞开用

 

 

 

③ DeepSeek V4 Flash · 深度推理之王

 

   256K 超长上下文,专治复杂逻辑推理、长文档分析、代码生成。DeepSeek 原厂出品,推理能力在免费阵营里属于最强一档。  

     

💰 150-500 次 / 5 小时 · 省着用,留给难活

 

 

 🟢 Part 3:三种用法,从零到上手只需 3 分钟

 

 ⏰ 本部分阅读约 2 分钟

 

 先注册:浏览器打开 token.sensenova.cn,手机号收验证码登录 → 首页弹窗点「领取公测包」→ 价格 0 元/月 → 左侧菜单「API Key 管理」创建密钥,复制保存(只显示一次!)。

 

     

🟢 方式A:网页直接聊(零门槛)

       

登录后首页就是对话窗口,选模型→打字→回车。传文件点📎图标。不需要装任何东西。

 

 

     

🟠 方式B:ChatNextWeb 客户端(推荐)

       

GitHub 搜 ChatNextWeb 下载安装,配置一次永久用,支持多模型切换对比、历史记录、语音输入。

         

       Base URL:https://token.sensenova.cn/v1
       API Key:粘贴你注册后获得的 sk- 开头字符串      

     

 

     

🔵 方式C:API 代码调用(最灵活)

       

完全兼容 OpenAI 格式,会调 ChatGPT 就会调日日新:

         

from openai import OpenAI
client = OpenAI(
    api_key="sk-你的密钥",         # 从商汤后台获取
    base_url="https://token.sensenova.cn/v1"
)
response = client.chat.completions.create(
    model="SenseNova 6.7 Flash-Lite",
    messages=[{"role": "user", "content": "帮我写周报..."}]
)

     

 

 🟠 Part 4:Agnes AI —— 文本+生图+视频,永久免费!

 

 ⏰ 本部分阅读约 2 分钟

 

 Agnes AI(agnes-ai.com)——2026 年 6 月起三大核心能力无限期免费。不是公测、不是限免,是真的永久。

 

     

💬 agnes-2.0-flash · 文本对话

   

日常问答、写文章、翻译、写代码,替代 ChatGPT。后端 vLLM 引擎,响应速度很快。

       

🎨 agnes-image-2.0-flash · AI 生图

   

输入描述词直接出图,1024×1024 分辨率。2.0 版无水印。返回公网可访问的 PNG 链接。

       

🎬 agnes-video-v2.0 · AI 生成视频

   

输入一句话描述,自动出几秒视频片段。异步任务,提交后等 1-3 分钟取结果。

 

 

 

   一句话:你平时用 AI 干的三件事——聊天、画图、做视频,一个平台全包,全部免费,永久。  

 

📋 Agnes AI 接入配置(三样东西)

 

                                                               

Base URL https://apihub.agnes-ai.com/v1
API Key 去 agnes-ai.com 注册获取你自己的密钥
协议 完全兼容 OpenAI,即插即用

 

 

文本调用示例

 

   model: "agnes-2.0-flash"
   格式: OpenAI Chat Completions 标准格式
   直接替换 api_key 和 base_url 即可  

 

 

生图调用示例

 

   model: "agnes-image-2.0-flash"
   size: "1024x1024"
   接口: /v1/images/generations  

 

 

视频生成示例

 

   model: "agnes-video-v2.0"
   接口: /v1/videos(异步,先提交后查询)
   参数: prompt + frames + fps  

 

 ⚪ Part 5:两平台加起来,这 6 个坑别踩

 

 ⏰ 本部分阅读约 2 分钟

 

 商汤 + Agnes,实测踩过的坑都在这里了。

 

🟠 商汤 3 个坑:

 

 

❌ 坑1:5小时重置,不是每天

 

从你第一次调用开始计时,5小时后恢复。上午10点用完,下午3点才回血,不是凌晨重置。

 

 

❌ 坑2:公测可能后续收费

 

趁现在免费注册锁定身份,早注册不吃亏。Agnes AI 已确认永久免费可以放心。

 

 

❌ 坑3:API Key 只显示一次

 

创建密钥后立刻复制保存到微信「文件传输助手」,关了页面就再也看不到了。

 

🔴 Agnes 3 个坑:

 

 

❌ 坑4:域名别搞错

 

API 网关是 apihub.agnes-ai.com,不是 api.agnes-ai.com。搞错连不上。

 

 

❌ 坑5:模型名必须带版本号

 

必须是 agnes-2.0-flash、agnes-image-2.0-flash、agnes-video-v2.0,填老版本名会 404。

 

 

❌ 坑6:视频是异步的,别傻等

 

提交视频任务后拿 task_id,每隔 2-3 分钟查一次结果。免费档有排队,耐心点。

 

 🎯 Part 6:白嫖党最优使用策略

 

 

❌ 错误:只用一家平台,用完就停了

 

   很多人注册了商汤,1500 次/5小时用完了就以为没了。其实 Agnes 是永久免费的,不知道有第二个平台。  

 

 

✅ 正确:双平台组合拳

 

   主力日常 → 商汤 Flash-Lite(1500次/5h,量大管饱)
   深度推理 → 商汤 DeepSeek V4(省着用,留给难活)
   AI 生图 → Agnes image-2.0(永久免费,无水印)
   AI 视频 → Agnes video-v2.0(异步提交,排队等待)
   长文本对话 → Agnes 2.0-flash(永久免费备用)  

 

 

❌ 错误:用 7B 模型做复杂推理,然后吐槽它弱

 

   很多新手用 Flash-Lite 去写长篇代码,做不好就说「免费的果然不行」。免费模型有明确的能力边界。  

 

 

✅ 正确:梯度使用法

 

   小活(翻译/摘要/分类)→ Flash-Lite
   中活(写稿/分析/代码)→ Agnes 2.0-flash
   大活(复杂推理/长文分析)→ DeepSeek V4  

 

 

💰 算一笔账

 

   
     

ChatGPT

     

¥160/月

     

仅文本

   
   
 
   
     

Claude

     

¥160/月

     

仅文本

   
   
 
   
     

Midjourney

     

¥200+/月

     

仅生图

   
   
 
   
     

商汤+Agnes

     

¥0

     

文本+生图+视频

   
 

 

   两套方案,六款模型:写稿 + 生图 + 做视频,全免费。
   一年省下四五千块工具费,省到就是赚到。  

 

 

   🚀 觉得有用?转发给身边的朋友,让大家一起告别 AI 工具月费!  

 

   
     

token.sensenova.cn

     

商汤 · 公测免费

   
   
     

agnes-ai.com

     

Agnes · 永久免费

   
 

 

 

 

 

💬 今日互动话题

 

   你用过商汤或 Agnes 吗?体验怎么样?或者你还发现了哪些免费好用的 AI 平台?欢迎留言分享,我会精选 3 条最有价值的留言,送出《免费 AI 工具百宝箱》整理文档。  

 

 

🎁 转发福利:转发本文到朋友圈/技术群,截图发到公众号后台,领取《2026免费AI工具全景图》高清PDF

 

 

📅 下期预告:周五 | Prompt工程课

 

《Prompt工程实战:3个模板让你的免费模型效果翻倍》—— 同样的免费模型,换个提示词方式,输出质量天差地别。

 

 #免费AI  #白嫖攻略  #商汤日日新  #AgnesAI  #AI工具

 

 本文信息基于 2026 年 6 月各平台公开信息整理。免费政策可能随时调整,请以各平台最新官方公告为准。

 

 

💰 Vol.004 · Free-Use Guide · 2026.06.25 ⏱ Read time: about 7 min

📍 Reading progress: Start exploring

🔍 SenseTime 🤖 Agnes 🛠 Hands-on ⚠️ Pitfalls 🎯 Summary

😱 AI tool costs are getting more and more expensive. ChatGPT costs ¥160/month, Claude costs ¥160/month, Midjourney costs ¥200+/month. People trying to start an AI side hustle haven't even earned a cent yet, but they've already paid several hundred in tool fees.

💡 But last week I found two real treasure platforms — SenseTime's SenseNova and Agnes AI. Together, the two platforms offer six large models covering writing, image generation, and video creation — all completely free.

⚡ One-sentence positioning: This article is the most detailed "SenseTime + Agnes dual-platform free-use guide" on the internet — from registration to API calls, from pitfall avoidance to cost savings. You can get started right after reading it. All information is based on real testing conducted in June 2026.

🔵 Part 1: Six free models, clear at a glance

⏰ Reading time for this section: about 1 minute

In one sentence: These six models cover everything from writing, translation, image generation, and data analysis to short video generation — covering all daily AI scenarios. The key point: you don't spend a single cent.

🔴 Part 2: SenseTime SenseNova — all-around utility player + deep reasoning

⏰ Reading time for this section: about 3 minutes

SenseTime offers three models, each billed independently — using up one model's quota doesn't deduct from another's.

① Flash-Lite · multimodal all-around utility player

It can read documents, look at images, recognize tables, and perform OCR analysis. It supports a 252K context window — you can throw an entire novel in at once and it can still understand it. It's the absolute workhorse for daily office tasks.

💰 1,500 calls / 5 hours · use freely

② U1 Fast · infographic/poster creator

It specializes in infographics, posters, and data visualization. Feed in data and it directly outputs bar charts/pie charts/flowcharts, and it can even generate PPT-level layouts. Even if you don't know design, you can instantly become someone who "can make graphics."

💰 1,500 calls / 5 hours · use freely

③ DeepSeek V4 Flash · the king of deep reasoning

With a 256K ultra-long context window, it specializes in complex logical reasoning, long-document analysis, and code generation. Produced by DeepSeek itself, its reasoning ability is among the strongest in the free tier.

💰 150–500 calls / 5 hours · use sparingly, save it for the hard jobs

🟢 Part 3: Three ways to use it — from zero to proficient in just 3 minutes

⏰ Reading time for this section: about 2 minutes

First, register: open token.sensenova.cn in your browser, log in with your phone number and a verification code → click "Claim public beta pack" in the homepage popup → price: ¥0/month → go to the left menu "API Key Management" to create a key, then copy and save it (it's only shown once!).

🟢 Method A: Chat directly on the web (zero barrier)

After logging in, the homepage is the chat window. Select a model → type → press Enter. To upload files, click the 📎 icon. No installation needed.

🟠 Method B: ChatNextWeb client (recommended)

Search for ChatNextWeb on GitHub, download and install it. Configure once and use it forever. It supports multi-model switching and comparison, history records, and voice input.

Base URL: https://token.sensenova.cn/v1 API Key: paste the string starting with sk- that you obtained after registration

🔵 Method C: API code calling (most flexible)

It is fully compatible with the OpenAI format. If you can call ChatGPT, you can call SenseNova:

🟠 Part 4: Agnes AI — text + image generation + video, free forever!

⏰ Reading time for this section: about 2 minutes

Agnes AI (agnes-ai.com) — since June 2026, its three core capabilities have been free with no time limit. Not a public beta, not a limited-time offer — truly permanent.

💬 agnes-2.0-flash · text chat

For everyday Q&A, article writing, translation, and coding, it can replace ChatGPT. Powered by a vLLM engine on the backend, with very fast response times.

🎨 agnes-image-2.0-flash · AI image generation

Enter a description and it directly generates an image at 1024×1024 resolution. Version 2.0 has no watermark. It returns a publicly accessible PNG link.

🎬 agnes-video-v2.0 · AI video generation

Enter a one-sentence description and it automatically generates a several-second video clip. It's an asynchronous task — after submitting, wait 1–3 minutes to retrieve the result.

In one sentence: The three things you normally use AI for — chatting, generating images, and making videos — are all covered on one platform, all free, forever.

📋 Agnes AI integration configuration (three things)

Text API call example

model: "agnes-2.0-flash" format: standard OpenAI Chat Completions format — just replace the api_key and base_url

Image generation call example

model: "agnes-image-2.0-flash" size: "1024x1024" endpoint: /v1/images/generations

Video generation example

model: "agnes-video-v2.0" endpoint: /v1/videos (asynchronous — submit first, then poll) parameters: prompt + frames + fps

⚪ Part 5: Combined across both platforms — these 6 pitfalls to avoid

⏰ Reading time for this section: about 2 minutes

SenseTime + Agnes: all the pitfalls we hit during real testing are right here.

🟠 SenseTime's 3 pitfalls:

❌ Pitfall 1: The 5-hour reset, not daily

The timer starts from your first API call, and the quota restores after 5 hours. If you use it up at 10 AM, it won't recover until 3 PM — it does not reset at midnight.

❌ Pitfall 2: The public beta may charge later

Take advantage of the current free registration to lock in your identity now. Registering early never hurts. Agnes AI has already confirmed it is permanently free, so you can rest easy.

❌ Pitfall 3: The API Key is only shown once

Immediately copy and save the key to WeChat's "File Transfer Helper" after creating it. Once you close the page, you can never see it again.

🔴 Agnes's 3 pitfalls:

❌ Pitfall 4: Don't get the domain wrong

The API gateway is apihub.agnes-ai.com, not api.agnes-ai.com. Get it wrong and you can't connect.

❌ Pitfall 5: Model names must include the version number

It must be agnes-2.0-flash, agnes-image-2.0-flash, or agnes-video-v2.0. Using old version names will return a 404.

❌ Pitfall 6: Video generation is asynchronous — don't just sit and wait

After submitting a video task, grab the task_id and poll for results every 2–3 minutes. The free tier has queueing, so be patient.

🎯 Part 6: Optimal strategy for free users

❌ Wrong: Use only one platform, then stop once it runs out

Many people register for SenseTime, use up the 1,500 calls per 5 hours, and think it's over. But Agnes is permanently free — they just don't know about the second platform.

✅ Right: A combined two-platform strategy

Main daily driver → SenseTime Flash-Lite (1,500 calls/5h, plenty of volume) Deep reasoning → SenseTime DeepSeek V4 (use sparingly, save it for hard jobs) AI image generation → Agnes image-2.0 (permanently free, no watermark) AI video → Agnes video-v2.0 (submit asynchronously, wait in queue) Long-form text chat → Agnes 2.0-flash (permanently free backup)

❌ Wrong: Using a 7B model for complex reasoning, then complaining it's weak

Many beginners use Flash-Lite to write long-form code, and when it doesn't work out, they say "the free ones really don't work." Free models have clear capability boundaries.

✅ Right: The tiered usage method

Small tasks (translation/summarization/classification) → Flash-Lite Medium tasks (writing/analysis/coding) → Agnes 2.0-flash Heavy tasks (complex reasoning/long-document analysis) → DeepSeek V4

💰 Let's do the math

ChatGPT

¥160/month

text only

Claude

¥160/month

text only

Midjourney

¥200+/month

image generation only

SenseTime + Agnes

¥0

text + image generation + video

Two plans, six models: writing + image generation + video creation, all free. You'll save four to five thousand yuan a year in tool fees. Saving money is making money.

🚀 Found this useful? Share it with your friends and let everyone say goodbye to monthly AI tool fees!

token.sensenova.cn

SenseTime · free public beta

agnes-ai.com

Agnes · permanently free

💬 Today's interaction topic

Have you used SenseTime or Agnes? How was the experience? Or have you discovered other free and useful AI platforms? Feel free to leave a comment. I'll pick the 3 most valuable comments and give away the "Free AI Tool Treasure Box" compiled document.

🎁 Sharing reward: Share this article to your Moments/tech group, send a screenshot to the official account backend, and get the high-definition PDF "2026 Free AI Tool Landscape Map"

📅 Next issue preview: Friday | Prompt Engineering Lesson

"Prompt Engineering in Practice: 3 templates to double the performance of your free models" — With the same free model, simply changing how you phrase your prompt can lead to wildly different output quality.

#FreeAI #FreeUseGuide #SenseTimeSenseNova #AgnesAI #AITools

The information in this article is compiled from publicly available platform information as of June 2026. Free policies may be adjusted at any time. Please refer to each platform's latest official announcements.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry