Developers are going crazy! The world's top 10 AI labs are now completely free to use, burning through 3.12 trillion tokens in just one week.

The carnival of full-modal computing power has begun: The top ten global AI giants are offering free APIs indefinitely, with the weekly call volume exploding to 3.12 trillion Tokens! This week, Agnes has made a game-changing upgrade: 1M ultra-long context + 4K ultra-clear image quality can be obtained "for free". The open-source community has gone crazy. Independent developers and small teams, come and take advantage of it quickly!

In the past two weeks, a strange local inflation has been occurring in the AI circle.

All the large models are competing fiercely, and developers are gasping at the bills in the background.

In this era of full-scale Agents, the Token consumption is extremely high. Before the product is even in public beta, the bill can already break you.

Although all AI giants are burning Tokens, few are offering free APIs.

Against this background, on June 1st, Agnes AI, a top AI Lab ranked in the top ten of the global model list, announced that it would offer its core full-modal model APIs for free to developers and creators worldwide indefinitely.

This includes:

- Text model: Agnes-2.0-Flash

- Image model: Agnes-Image-2.1-Flash

- Video model: Agnes-Video-2.0

Note that this is not a limited-time trial. It's truly free.

After this news was announced, users' enthusiasm was instantly ignited.

Some people also questioned: Is it really so generous? Are they just using a weakened version of the model to deceive registrations?

Just half a month later, Agnes AI released the latest statistics for the second week, completely shattering all the doubts!

3.12T weekly total call volume. Who is going crazy for freebies?

This time, the call data of Agnes' full-modal models has once again refreshed the historical record: The weekly call volume of the total full-modal Tokens has soared to 3.12T (trillion)!

Among them, the text model (Agnes-2.0-Flash) alone contributed about 1.9T of the call volume.

The image and video models (Agnes-Image-2.1-Flash + Agnes-Video-2.0) together contributed up to 1.2T of the call volume.

The weekly call volume of 3.12T completely surpasses that of Claude Opus 4.7 on OPenRouter and is comparable to the top 5!

You know, in the industry, visual APIs have always been a luxury. When Agnes made its visual models free, the call volume in this area skyrocketed.

It directly reflects that the full-modal free policy has instantly liberated the productivity of visual content creation and batch creative testing.

These creators and developers, who have been held back by high costs, are using trillions of Tokens to prove that it's not that they don't have ideas for full-modal innovation. They just couldn't afford it before.

Lower price without sacrificing quality: Saving money while being "truly powerful"

In the AI circle, free and cheap often lead people to think of low quality.

But the reason why Agnes AI has won over the open-source community is that while saving money, it has pushed its underlying capabilities to the global forefront.

Let's take a look at a hardcore evaluation and pricing comparison table:

As an AI Lab ranked in the top ten of the global model list, Agnes' core logic is very straightforward: Save money without sacrificing capabilities.

Even before it was free, its price was only about half of that of similar mainstream commercial models. Now, when the full-modal matrix has zeroed out the price, it has directly achieved a dimensionality reduction strike!

This week's upgrade: 1M context goes live in a gray release + 4K ultra-high definition debut

If the free strategy in the previous two weeks was like giving out benefits, then the product upgrade that Agnes AI is about to complete this Wednesday is like dropping a bombshell.

Without changing any existing code and without spending an extra cent, Agnes has directly unlocked two advanced model capabilities for developers under the existing free framework.

Upgrade point A: The 1M native ultra-long context of Agnes-2.0-Flash will be implemented

This week, the text model will natively support an ultra-long context window of 1M Tokens. 50% has been released in a gray release and will be fully updated soon.

At the API level, there is no need for any complex configuration changes. As long as the total content of the messages array in the API request is within 1M Tokens, you can directly enjoy this million-level ultra-long memory for free.

The core value of the 1M context is not just about being able to fit more words. Its essence is to eliminate the information gaps and high development costs caused by document segmentation, slicing, and repeated context passing.

In this 1M "big pocket", the following scenarios, which were originally considered luxurious, have now become free standard features.

1. True understanding of code libraries and projects

Previously, when you asked AI to help you fix bugs, you could only paste code piece by piece. Now, you can package the source code, configuration files, dependencies, and project documents of an entire medium to large software project and send them to Agnes-2.0-Flash at once.

It can help you see the entire project's architectural relationships at a glance, conduct a global code review, and even directly locate hidden vulnerabilities in cross-file calls.

2. "Slice-free" reading of long documents and complex materials

Whether it's a whole long novel, a technical manual of hundreds of thousands of words for a large device, a complex legal contract, or multiple related scientific research papers, you no longer need to spend time building complex RAG slicing algorithms.

Just throw it in and let it answer detailed questions about the whole book and analyze cross-chapter information associations.

3. Long-term memory for long conversations of Agents

This solves the pain point that enterprise customer service and virtual assistants suddenly "forget" or deviate from their character settings after 100 rounds of conversation.

Upgrade point B: The 4K ultra-high definition text-to-image generation of Agnes-Image-2.1-Flash is fully unlocked

Previously, in the image generation API, a 1K resolution was the norm. If you wanted high definition, you had to pay extra or run other super-resolution models in the background.

After this upgrade, Agnes-Image-2.1-Flash has directly unlocked the ability to output 4K (up to 4096×4096) ultra-high definition images!

Moreover, it natively supports almost all mainstream aspect ratios on the market: 1:1, 3:4, 4:3, 16:9, 9:16, 2:3, 3:2, 21:9.

It can easily cover everything from e-commerce main images, self-media covers, and product posters to local retouches.

This directly unlocks more possibilities.

For example, turn a blurry image into a 4K high-definition large image with just one sentence, and generate a brand-new high-definition scene from an existing image:

An ultra-wide aerial panoramic view of the New York City skyline in Manhattan, taken from a high-rise observation deck. The Empire State Building is located slightly to the right of the center of the frame. The One World Trade Center can be seen in the distance. The Lower Manhattan skyline stretches to the horizon. The foreground is filled with dense skyscrapers and historical high-rise buildings. The details of the building windows and roofs are clear. The background has a light blue atmospheric haze. The space has rich layers. The sky is a clear light blue with a few thin cirrus clouds and airplane contrails. The natural sunlight is bright. The color tone is a cold blue like in a movie. It's a real architectural photography with a wide-angle lens, deep depth of field, HDR, high sharpness, photo-realistic quality, and professional urban landscape photography. 4K ultra-high definition with extreme details.

The night view of a futuristic city with neon lights on high-rise buildings, reflective streets after rain, movie-level lighting, ultra-clear details, 4K, and high quality.

The native 4K gives e-commerce main images the solemnity and poetry of a movie-level advertisement, making shopping full of a sense of texture and order.

After the upgrade of Agnes-Image, it not only improves the clarity but also directly enhances the detail and texture.

The most amazing thing is that its access method is extremely smooth.

Your original request for a 1K image looked like this:

Now, if you want to experience the high-definition details and delicate textures of 4K, you just need to change "size": "1K" to "size": "4K".

You don't need to change the code. The response still supports URL links and b64_json. The most important thing is that the cost of generating a 4K image is exactly the same as that of a 1K image - still zero.

This is a real liberation of productivity for designers and startup teams with needs for batch image generation, e-commerce scene replacement, and high-quality material testing.

This Friday, the full-modal voice link will be completed!

In addition, Agnes has also released important news in advance: It is expected that around this Friday (June 19th), the TTS (Text-to-Speech) ability will officially start a gray test.

The first version will offer 20 high-quality voices, covering different genders, age groups, and styles, and support both Chinese and English generation.

For developers and creators, this means that a more complete voice creation ability is becoming within reach.

The completion of this piece of the puzzle also marks the official closure of Agnes' full-modal automation loop - from text to images, from videos to voice, the entire link is finally connected.

You can use the large text model to write a script, use the image model to break down the storyboard, use the video model to directly generate a picture with sound effects, and finally use TTS to add the appropriate human voice narration. An entire AI content production line no longer needs to be pieced together. The whole process can be completed in one stop in Agnes' system