Token 无限免费,终于来了。Infinite Free Tokens, Finally Arrived.


AI free token/AI免费积分

 

上线两周,调用量 3.12 万亿 Token。

第三周的时候,用量已经达到了4.11万亿Token。

图片生成累计上千万张,视频生成累计数百万秒。

这些数据来自一个叫 Agnes AI 的产品。它是全球榜单前十的 AI Lab 。

6 月 1 日,他们搞了个大事情。

把文本、图片和视频这三个模型的 API 全部都免费对外开放了。

无限期免费开放三个模型的 API:

  • 文本模型 Agnes-2.0-Flash
  • 图片模型 Agnes-Image-2.1-Flash
  • 视频模型 Agnes-Video-V2.0

不是试用期,也没有上限,直接把调用费用降到0。

我先用 3个 案例给你们看看

文本模型:以前处理长文档要算 Token

一个搞运维的要对一份有 50 多页的硬件采购合同进行分析,并从中发现其中存在的风险条款。

图片

你要用GPT,光输入就消耗掉几万多个 token,再加上合同数据一条条地分析、生成报告,来回几次账单就会很高。

用上 Agnes-2.0-Flash:把整份合同丢进去,让它逐条分析并标注出风险点,还可以任意提问,费用全免。

图片

图片模型:以前做设计要算张数

一个老板要给新开的烧烤店做几个不同尺寸的海报,要发布到微信、微博、抖音这三个平台上。

很多人说用 Midjourney 来生成图片,效果好,单一张图大概要几十块钱,如果尺寸不对还要再生成一次。真心用不起。

用 Agnes-Image-2.1-Flash:给它一句话,然后让其产生多种大小的海报,要横版有横版,要竖版有竖版,全免费。

图片
图片

看一下成片,很有感觉吧。

图片

视频模型:以前做素材要算秒数

我用Agnes-Video-2.0做了一个15秒的视频。

图片

整个过程大概就 5 分钟左右。成片的质量还可以,大家自己看。

Token 无限免费,终于来了。
已关注
关注
重播 分享 赞
 
开源日记
 
 
 

 

 

0/0

00:00/00:15
 
 
进度条,百分之0
 
 
00:00
/
00:15
00:15
倍速
全屏

 

倍速播放中
 
 

继续观看

Token 无限免费,终于来了。

 
转载
,
Token 无限免费,终于来了。
开源日记
已同步到看一看写下你的评论
 

 

 
 
视频详情

这要用 Runway 来做,每秒视频几毛钱,光是生成费用就要一两百块。要是效果不理想的话,还要再做一遍。

它把三种全模态模型的能力都给你

01 文本模型:一句话就可以替代一本书

Agnes-2.0-Flash 最多可以支持 1M 的上下文,但是在调用高峰的时候会被限制在 512K。

1M 这到底意味着什么?

可以把一本书、一份合同或者整个项目的全部代码扔进去,不需要切片、不需要分段,它会自己去理解。

以前做长文档分析要不断地分段、合并,现在可以直接把整个文档交给模型来处理。

02 图片模型:4K超清、可修改

Agnes-Image-2.1-Flash 可以生成 4096×4096 的 4K 图片,比一般的高清还要清晰一点。

不仅可以生成图片,还可以对图片进行修改、换背景、局部修改、变风格。

03 视频模型:画面声音一起出

视频生成本来就是个烧钱的事情,现在也可以免费用了。

Agnes-Video-V2.0 可以做到画面和声音同时出现,并且不需要人工配音。可以用一张图来生成视频,用两张图做过渡,做出多个镜头的内容。

免费不代表实力减弱,在评测数据中可以看得出是有底气的

很多人一听到“免费”,第一反应就是:

免费的东西,能力一定不好吧?

但是从公开的数据来看,Agnes 还是很能打的。

Agnes-2.0-Flash 在全球 Agent 能力榜上排名前十,在部分任务中甚至超过了Gemini Flash 和 MiniMax M2.7。

Agnes-Image-2.0-Flash 也进入了 Artificial Analysis 图片编辑榜的世界前列。

图片

注意 Artificial Analysis 采用的是真实用户的盲测方式,评判者不知道作品来自哪个模型,只根据结果打分。

同时成品数量也创新高。

图片模型每周能产生超过 567 万张的图片,视频模型每天能够产生超过 237 万秒的视频。

图片、视频占的比例很大,也说明免费政策确实击中了用户需求。

所以 Agnes 想证明的并不是“免费但缩水”,而是:

同样的品质,更低的门槛,甚至直接免费。

而且它兼容主流工具,上手成本几乎为零

Agnes 接口与 OpenAI 完全兼容,只须改动一个地址即可使用。

怎样接入呢?非常简单:

首先到 platform.agnes-ai.com 上注册一个账号,并获得API key

然后打开 cc-switch,填上 https://apihub.agnes-ai.com/v1 和 上面的 API key

图片

模型 Key 别填错了。

图片

最后打开  ClaudeCode 就可以用了。

图片

文本、图片、视频三种模式都可以使用,并且调用也是一样的。

写在最后

注意目前视频生成是异步的,不能实时拿到结果,需要等一会儿再查询。

高峰期排队可能要等 3-5 分钟,不像文本和图片那样秒出。

生成出来的视频链接是暂时性的,要尽快下载并保存好,否则过了期限就没有了。

感兴趣的朋友们可以去 https://agnes-ai.com 试一试。

另外官方还开了一个 GitHub 仓库来追踪用户的反馈问题。

https://github.com/AgnesAI-Labs/Agnes-AI/issues

遇到响应慢、图片偶发报错、客户端接入失败这类问题,都可以去这里找答案。

图片

既然看到这了,欢迎随手点赞、在看、转发,也可以给我个星标⭐,接收最新的文章,我们下期见!

 

In two weeks, the call volume reached 3.12 trillion tokens.

By the third week, the usage had climbed to 4.11 trillion tokens.

Image generation accumulated over ten million pictures, and video generation accumulated millions of seconds.

These data come from a product called Agnes AI. It is one of the top ten AI labs in the global rankings.

On June 1st, they made a big announcement.

All three models' APIs — text, image, and video — were opened for free, with no restrictions.

Unlimited free access to the APIs of three models:

Text model Agnes-2.0-Flash

Image model Agnes-Image-2.1-Flash

Video model Agnes-Video-V2.0

This is not a trial period; there is no upper limit. The call fees have been reduced to zero.

Here are 3 examples for you to see.

Text model: Previously, processing long documents required paying for tokens.

An operations manager needs to analyze a hardware procurement contract with over 50 pages to identify any risky clauses.

If you use GPT, just inputting the data would consume tens of thousands of tokens. Plus, analyzing the contract line by line and generating a report would incur high costs.

With Agnes-2.0-Flash: You can throw the entire contract into the system, have it analyze each clause and mark the risks, and even ask follow-up questions — all for zero cost.

Image model: Previously, design work required paying per image.

A restaurant owner wants to create several posters for a new barbecue shop, in different sizes for WeChat, Weibo, and Douyin.

Many people suggest using Midjourney for image generation, which is high quality but costs about 50 yuan per image. If the size is wrong, you have to generate it again — very expensive.

With Agnes-Image-2.1-Flash: Give it one prompt, and it generates posters in various sizes — landscape, portrait — all free of charge.

Here’s the final product; the feel is very nice.

Video model: Previously, making clips required paying per second.

I used Agnes-Video-2.0 to make a 15-second video.

The whole process took about 5 minutes. The quality of the final clip is decent; you can see for yourself.

Already following

Follow

Replay Share Like

Close

Watch More

More

Exit Fullscreen

Open Source Diary (following)

Share video

, duration 00:15

0/0

00:00/00:15

Switch to horizontal mode

Continue playing

Progress bar, 0%

Play

00:00

00:15

00:15

Speed

Fullscreen

Speed playback in progress

0.5x 0.75x 1.0x 1.5x 2.0x

High definition Smooth

Continue watching

Tokens are now free, finally!

Watch more

Repost

,

Tokens are now free, finally!

Open Source Diary (following)

Share Like Thought

Synced to Glance Write your comment

Video details

If you used Runway for this, each second of video costs several cents — meaning the generation fee alone would be over a hundred yuan. And if the result isn’t ideal, you’d have to pay again.

It offers the capabilities of all three modal models:

01 Text model: One sentence replaces a book.

Agnes-2.0-Flash supports up to 1 million words of context (though limited to 512K during peak times).

What does 1M mean? You can feed in an entire book, contract, or full project code without slicing or segmenting — it understands the whole thing.

In the past, long-document analysis required continuous segmentation and merging; now you can hand over the entire document for processing.

02 Image model: 4K ultra-clear, editable.

Agnes-Image-2.1-Flash can generate 4096×4096 4K images — higher resolution than standard HD.

Not only can it generate images, but it can also modify them — change backgrounds, edit local regions, or apply different styles.

03 Video model: Audio and visuals in one go.

Video generation is usually expensive, but now it’s free too.

Agnes-Video-V2.0 produces both visuals and sound simultaneously, without needing manual voiceover. You can generate video from an image, create transitions between two images, or produce multi-camera content.

Free doesn’t mean compromised quality; the benchmark data confirm it’s competitive.

Many people assume “free” equals “low quality.” But the public data show Agnes is quite capable:

Agnes-2.0-Flash ranks in the top ten globally for agent capability, outperforming Gemini Flash and MiniMax M2.7 in some tasks.

Agnes-Image-2.0-Flash is also among the world leaders in the Artificial Analysis image editing leaderboard.

But looking at the public data, Agnes is still very capable.

Additionally, output volume is high: over 5.67 million images per week from the image model, and more than 2.37 million seconds of video per day from the video model.

The heavy use of images and videos suggests the free policy is meeting user demand.

So Agnes isn’t proving “cheap and shrunk” — it’s proving “same quality, lower barrier, even free.”

And it’s compatible with mainstream tools, requiring nearly zero learning curve.

Agnes’s API is fully compatible with OpenAI’s — just change the endpoint and you’re ready.

How to connect? Very simple:

Register at platform.agnes-ai.com and get your API key.

Open cc-switch, enter https://apihub.agnes-ai.com/v1 and your key.

Don’t confuse the model key.

Then open ClaudeCode and start using it.

All three modes (text, image, video) work through the same interface.

Final notes: Video generation is asynchronous — you won’t get the result instantly; you need to wait and query later.

During peak hours, there may be a 3–5 minute queue, unlike text and image which are nearly instantaneous.

Generated video links are temporary — download and save them quickly before they expire.

If interested, try it at https://agnes-ai.com.

There’s also a GitHub repo for tracking user feedback and issues: https://github.com/AgnesAI-Labs/Agnes-AI/issues

For slow responses, occasional image errors, or client integration failures, check there for answers.

Since you read this far, please like, share, and star ⭐ this post for future updates. See you next time!

图片

图片

图片

图片

图片

图片

Infinite Free Tokens, Finally Arrived.

图片

图片

图片

图片

图片

图片

Infinite Free Tokens, Finally Arrived.

图片

图片

图片

图片

图片

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