Suddenly, Sonnet 5.5 is running!

  It's the first time in the history of the world.

Who would have thought that Opus 5.5 would be released in a few days, and that Sonnet 5.5 would also be coming?

Since last night, the entire X has been spreading the message that Sonnet 5.5 will be released soon, probably on Tuesday afternoon US time.

In performance, the Sonnet 5.5 is closer to Opus 5.5, but much cheaper than Opus 5.5.

Now, Sonnet 5.5 is in the final stages of its release and has started gray testing in Claude Code.

Several demos revealed that the Sonnet 5.5 torque suppressed the newly released GPT-6 Sol, and even pushed OpenAI's flagship GPT-6 Astra in terms of coding and Agent capabilities!

What's even more shocking is that it's price is right in the heart of the country, down to $2 million in tokens.

If the timing of the outbreak is correct, it looks like Anthropic is poised to target Dev Dayl of OpenAI.

Recall that on September 22, when Anthriopic suddenly released Opus 5.5, a handwritten statement in the official document read: "Sonnet 5.5 and Haiku 5.5 will be released in the coming weeks".

No one would have thought Sonnet 5.5 would come so soon!

Just yesterday, developer V @Mr_Salio blasted on X: Claude Sonnet 5.5 has been captured!

In the artifact, a model configuration identifier for the Sonnet 5.5 exclusive has emerged:claude-sonnet-5-5」。

Not only that, but it's also in the same profile".claude-opus-5-5」、「claude-sonnet-5」And the mysterious,claude-fable-5.1」。

Writing the front-end code means that the model is already installed on the server, waiting for a switch.

Subsequently, big V @notjazii also blasted out that Sonnet 5 was actually routed to version 5.5 about a week ago for "assumption testing".

Have you been measured in ashes? This time it's the old way.

Open up your Claude Code, turn off the network and memory functions, and ask it directly: Do you know that reinstalling sibling tibo?

If you're using an older version of the model, it may be nonsense; but if you've been routed to version 5.5, it will tell you exactly what happened to Tibo and Codex.

Some netizens have tested verification that some accounts have indeed been silenced in the past, and they are experiencing Opus 5.5 unprecedentedly.

Remember to use Claude Code

To summarize the findings that have been shared so far, does the Sonnet 5.5 actually compress the GPT-6 Sol?

First, it's a pure JS front-end UI/animation boundary contrast, with four mainstream models competing on the same platform.

Developers @notjazii have released a series of highly visually impactful observations: "Four mainstream models generate contrast with pronouns".

He asked the model to produce complex UI animation effects.

@notjazii commented on the contrasts: "Sonnet 5.5 really rubs both models of OpenAI (Sol and Astra) on the ground!"

Indeed, as can be seen from the diagram below, Sonnet 5.5 exhibits a super-powerful coding intuition and aesthetic taste.

GPT-6 Astra

GPT-6 Sol

Claude Opus 5.5

Claude Sonnet 5.5

Some netizens were shocked to see the results: "Claude has always been stronger in UI and animation, but it's hard to believe it's true if it's pure code generated!"

Big V @chetaslua focuses on Sonnet 5.5's daily coding capabilities and code handles.

He used the same three prompts, running six parallel runs, each generating one model at a time, with output results unchanged.

Slide to the left and look at the picture

Sonnet 5.5 eventually generated three files of 90 tonnes131 KB, touching the 90-minute ceiling in all runs; GPT-6 Sol generated files between 20 tonnes29 KB in size and took 9-12 minutes.

He exclaimed, "Now go with Claude Code! Sonnet 5.5 in Ultracode mode is incredibly fast, not only extremely efficient, but it brings back the extremely natural 'human taste' of the Sonnet 4 era!"

In his view, at the same price, with the same hints, Sonnet 5.5 achieved torque pressure on the GPT-6 Sol.

In fact, what OpenAI fears is that Sonnet 5.5 is no longer as simple as pressing the GPT-6 Sol.

Following early testing by @srikanthvaluri and @buildwithrajath, the exploders pointed out: "Sonnet 5.5's performance far exceeded expectations, and while the GPT-6 Astra still has a slight edge in fineness, Sonnet 5.5's performance directly parallels Astra".

Unsurprisingly, in some advanced coding and Agent capabilities, it also approaches Astra.

Some people feel that even the Sonnet has crushed Astra, and what age are we entering?

If the excess of testing capabilities only makes OpenAI feel "uncomfortable", then the pricing of Sonnet 5.5 exposure is digging OpenAI's roots.

It was revealed that the performance of the "fairy model" of the Astra, which beat the Sol, was fully priced to the low-end market.

Input: $2 / million Token

Output: $10/million Token

Cache reads: only $0.20/million Token!

This means that Anthropic is delivering cutting-edge AI capabilities with "white price".

OpenAI's GPT-6 Sol and Luna are half the price of the GPT-5.6 and aim to capture the high-frequency scene.

But now, with the Sonnet 5.5, Anthropic hits the same price range, and the performance is much stronger.

"If Anthropic could deliver such a huge performance leap as Opus 5.5 at such a highly competitive price, OpenAI would be powerless".

The developer, Rajath Gowda, said frankly: "Sonnet 5.5 is becoming a big problem for OpenAI. If it's really close to Opus 5.5, Anthropic is going to make its front-end capabilities extremely cheap. That's what OpenAI is most worried about".

Anthropic's strategic ambition is clear: to defend the high-end throne with Opus 5.5 and take over the mid-end and everyday development market with Sonnet 5.5 at extreme cost.

Moreover, the choice of release time is often more dramatic than the product itself.

On the same day that OpenAI released the GPT-6 Sol and Luna, Anthropic directly unveiled the Claude Opus 5.5, which is faster than its predecessor, with 40% less cost and 30% faster.

In just two weeks, Anthropic is even bigger: Claude has calculated a world-record nine-circle magnitude in physics; discovered an unknown biosynthesis system; and launched the Claude Marketplace, which contains more than 2,000 plugins.

In contrast, OpenAI has not only been the subject of internal model controversy in recent years, but has also exposed a number of out-of-coordination incidents such as AI leaks of user images and openly researched GitHub Token scams.

Next Tuesday (29 September) is DevDay, the annual conference of developers that is highly regarded by OpenAI.

At this point, the Sonnet 5.5 grayscale brushes the entire network.

Many of the big boys were crazy: "The model is expected to be released next week, I guess it's next Tuesday (DevDay) "

Maybe there will be a big preheating tomorrow at 2pm New York time.

Some people have said, "I'm sorry about OpenAI. They'd better have a strong response on DevDay, or they're really fucked".

Perhaps the reason for this is that Anthropic has already set up a post-RL/training process.

The key to everything is high-quality data.

When asked how Anthropic was able to mutate so quickly, industry mogul @yacineMTB speculates that the core is extremely high standards of data flying.

"But I'm still shocked, how did they do it?"

Anthropic has developed a large-scale "inner loop" model that they are using.AIIt's about making AI.

According to the latest internal quantitative data released, Anthropic has shifted the "data flyer" to a turnover that would shock its opponents:

• 26% of core R&D work has been done byAI(Claude's model) dominated completion.

• Up to 80% of the self-produced code is written independently by large models.

It's not just as simple as "reduction in efficiency", which means technically.The return to self-improvement (RSI) is beginning to appear.。

Anthropic estimates that as early as 2027, Claude could be fully automated, possibly at a more insane pace.

As ASI is in the middle of a war, Google is in the middle of a war, and we, the average user, are just sitting around waiting for the war.

Big V Token Gremlin says, "Next week's AI circle will have a scene that we've never seen before. Take your time, clear your schedule, get mentally prepared, and save your hardest workflow for the future. Believe me, what's coming will completely change the way you work".

The arrival of Sonnet 5.5 marks the official arrival of "the age of high-quality AI".

For two dollars, you can buy a million tokens of "White Green Building".

A very low-cost caching reader ($0.20/M) makes it easy for anyone to build a personal AI assistant that can last longer in context memory.

If it's true, in less than 24 hours, we'll have a new model that we can make.