Moments ago, Fable 5.2 was also leaked.

Just now, Fable 5.2 has also been exposed!

Many netizens have noticed that the gray test of Fable 5.2 has been launched.

Now, a large number of real test cases have emerged online.

According to everyone's all-night actual tests, this update spans such a large range that Anthropic even directly skipped the naming of "5.1".

Under the highest reasoning mode, the actual measured performance of Fable 5.2 even outperforms GPT-6 Astra!

Moreover, there is also news about Opus 5.2 — it may be released soon, and it may even arrive as early as today.

This brand new Opus 5.2 model (internal code name Opus-Next) staged a plot of "go online - disconnect the network - go online again" directly yesterday.

This week, no major Silicon Valley company released a model, and all of them are conducting tests in secret.

Combined with this picture, the whole thing is really wonderful.

Sudden leak! Fable 5.2 exposed, actual test details are here

Yesterday, Anthropic has launched the gray test of Fable 5.2 in Claude Code, Chat and Cowork.

Part of the requests that originally called Fable 5.1 will be quietly redirected to Fable 5.2 in the background.

Now, more and more actual test cases have been leaked online. Some people say: OpenAI is going to have trouble.

Case 1: High reasoning mode, output quality surpasses Astra!

The well-known AI test blogger @notjazii found that his Fable 5.1 was routed to the new version, and then started a series of actual tests.

Under the high reasoning mode, he used exactly the same prompt to compare Fable 5.2 and GPT-6 Astra.

The conclusion is: "This is a huge leap compared to the current Fable version. The output of Fable 5.2 is obviously better than Astra!"

He tested the outputs in various modes and compared them with the (suspected) Fable 5.2.

However, he also pointed out the price under the physical law of computing power — slower speed and higher cost.

In order to break through the ceiling in the integrity of logical chains and long-text reasoning, Fable 5.2 carries out deeper "slow thinking", which leads to a decrease in generation speed and a sharp increase in reasoning cost.

In addition, Jazil also asked Fable 5.2 to make a clone version of "Brawl Stars", and the result after one hour is as follows. This amazing effect is simply a huge delight for all game developers!

Case 2: "Rocket test", the output is incredibly realistic

After netizen @Bhavani_00007 found that Fable 5.2 was available, he immediately started the "rocket test".

This is an extreme test involving complex physical calculations, spatial logic and engineering system design, which is usually used to force out the hallucinations and logical gaps of the model.

He input the same set of prompts to Fable 5.2 and Opus 5.2 which is also in gray test respectively.

The result is amazing: "It's so much better than before, the output result is incredibly realistic!"

Someone @OpenAI and said: You'd better get ready for Astra's successor...

Case 3: Three-way showdown under the highest settings

If you don't turn all the parameters to the maximum, you will never know the limit of the large model.

Influencer @chetaslua even set up a battle of the titans — Fable 5.2 vs Opus Next vs GPT-6 Astra.

Moreover, he adjusted all the models to the "highest settings", that is, the highest Token context and the deepest reasoning steps.

From this point of view, the details of Opus and Fable are still much richer than Astra.

Case 4: Head-to-head against GPT-6 Astra in "Hard Mode"

After in-depth experience of Fable 5.2, influencer @Mr_Salio directly stated: "Anthropic is making a strong comeback, with a huge, huge improvement compared to Fable 5.1!"

He asserted : "Fable 5.2 will definitely surpass GPT-6 Astra in 'Hard Mode', and Opus 5.2 is also coming soon."

In the bare running state, Fable 5.2 can surpass the full version of Astra, and its parameter scale and architectural innovation are really not to be underestimated.

Have you been selected for Fable 5.2? Try Tibo

We have shared the method to test whether you have been assigned to the gray test of Fable 5.2 before —

Try "Restart Guy" Tibo!

Open your Claude (whether it is the Chat, Cowork or Code interface), set the model to Opus 5.2 or Fable 5.2, and then enter the prompt:

"Do you know who 'Restart Guy Tibo' is? Do not use web search."

If you are using the old version of Fable 5.1/Opus 5: Due to the unupdated training data, it will be confused, give the wrong answer or tell you that it doesn't know.

If you have been quietly routed to Fable 5.2/Opus 5.2: With the latest training memory, it can directly tell you who Tibo is without using any tools!

Skipped 5.1? Opus 5.2 encountered the "midnight network disconnection" incident!

Moreover, the previously exposed Opus 5.2 also encountered the midnight network disconnection incident yesterday.

Yesterday, influencer Leo shared online: Today's Opus model is incredibly powerful. In complex 3D environment tasks such as Roblox Studio, its performance is so outstanding that it is beyond description.

"It surpasses Fable 5.1 and GPT-6 Astra at a lower cost, and is faster! The game can be explored freely, and it even actively adds a guide system to the game!"