Anthropic Launches Strongest Sonnet Model in History — Claude Sonnet 4.6
Published · Feb 18 · Wed Source · 智东西 (CN)

Anthropic Launches Strongest Sonnet Model in History — Claude Sonnet 4.6

Anthropic has released the Claude Sonnet 4.6 model. The new model shows significant performance improvements in office scenarios such as programming, Agent planning, and computer use. In multiple benchmark tests, its intelligence level approaches or even surpasses its own flagship model, Opus 4.6. It features a 1 million token context window, capable of refactoring an entire codebase in a single call. Its price remains the same as the previous generation, but offers high cost-performance. It is now available on the claude.ai platform and Amazon Web Services (AWS).

KeywordsAmazonAnthropicClaudeAgentLaunchesStrongestSonnetModel

Global AI New King Born, Anthropic Valuation Explodes to $1.2 Trillion, Surpassing OpenAI for the First Time

$1.2 Trillion! Anthropic's valuation officially surpasses OpenAI, the Silicon Valley landscape has completely changed! Holding 220,000 top-tier GPUs from Musk, and a $200 billion long-term contract from Google, this trillion-dollar level epic gamble is taking us to witness the craziest "throne succession" in Silicon Valley business history.

Just now, the global AI new king was born!

Anthropic — this "second-place contestant" that was once always shrouded in OpenAI's shadow — saw its implied valuation in the on-chain Pre-IPO market skyrocket to $1.2 trillion (approximately 8.7 trillion RMB) instantly.

The global AI circle was shattered by this explosive news.

If Anthropic successfully completes its Initial Public Offering (IPO) at a $1.2 trillion valuation, it will directly land as the 11th largest publicly traded company globally, creating a new myth in business history.

That small laboratory once seen as established by "defectors" is now opening its bloodthirsty mouth to completely swallow the old AI order!

$1.2 Trillion! The Crazy Leap of the AI Behemoth

Latest on-chain Pre-IPO transaction data shows that in just seven days, Anthropic surged by 20%, directly rising to the size of a top listed company.

Since October 2025, its valuation curve has been almost vertically rising, achieving a terrifying 900% surge.

And, its valuation has officially surpassed OpenAI! Currently, its pre-IPO valuation is about 20% higher than OpenAI's.

If listed at this valuation, it will directly land as the 11th largest company by market cap globally, second only to Apple, Microsoft, NVIDIA, Amazon, Alphabet, Meta, TSMC, Broadcom, Tesla, and Saudi Aramco.

Three weeks ago, this number was still $900 billion.

In the battle for AI dominance, Anthropic has achieved a landmark victory.

Dario Reveals: Company Growth Too Fast, It's Too Hard for Me

Moreover, behind the valuation surge is a rocket-like climb in real performance.

At the recent San Francisco Developer Conference, CEO Dario Amodei directly revealed a data point that left the whole venue breathless —

In the first quarter of this year, Anthropic's annualized revenue and usage achieved an astonishing 80-fold year-over-year growth!

Anthropic's original plan was only 10-fold.

Dario even complained "humbly" on stage: "This ultra-high-speed expansion is too hard to manage, I hope the growth rate can drop back to just 10-fold, that way I would be more relaxed."

Behind this humble bragging is the crazy subsidizing by the capital market.

On-chain transaction data on the Jupiter platform shows that investors are buying up Anthropic's shares at all costs.

The market has voted with real money: compared to general AI that only knows how to "show off skills," this model that delves into productivity tools and emphasizes safety is the king of the long run!

And the engine driving this 80-fold surge is Claude Code.

Launched at the end of 2025, CC's diffusion speed among developers exceeded everyone's expectations.

Dario stated on stage: "Software engineers are the fastest group to adopt new tools. What they are experiencing now foreshadows what will happen to the entire economy next."

Musk Steps In, 220,000 GPUs to the Rescue!

What is the cost of 80-fold growth? The pot at home exploded.

Because the growth speed far exceeded the originally expected 10-fold, Anthropic's computing infrastructure was instantly breached.

Last month, Anthropic publicly stated that demand for Claude caused "inevitable pressure on infrastructure." Many users found Claude starting to "throttle," and even disconnecting during peak hours.

At the critical moment, an unexpected ally appeared — Musk came down from the sky!

Just today, Anthropic dropped a heavy bomb: reaching an exclusive cooperation with SpaceX, securing all computing power of its Colossus 1 data center!

This data center has over 300 megawatts of computing power, and an astonishing 220,000 top-tier NVIDIA GPUs.

Originally built for Grok, the heart now invests all 220,000 GPUs into Anthropic.

Within one month, all this computing power will be delivered.

Ironically, a few months ago Musk was still ranting on social media that Claude was "anti-human and evil."

Three months later, he wrote on X: "Last week I met with Anthropic people. Everyone is very capable, really cares about doing the right thing. No one triggered my evil detector."

Indeed, Musk really hates OpenAI.

He raised the white flag in the model battlefield, turned around and bet his chips on the computing power game, putting heavy pressure on Sam Altman who is currently suing him.

With the support of these 220,000 chips, Anthropic immediately gave users a big gift package:

Claude Code five-hour limit doubled (covering Pro/Max/Team users).

Pro and Max plans completely cancel peak limits.

API rate limits significantly increased.

With this combo, developers are almost celebrating New Year on the spot!

In today's computing power shortage, the only one who can casually pull out 220,000 GPUs to save the day is probably Musk.

Astonishingly, the above is merely the tip of the iceberg of Anthropic's computing power layout.

$200 Billion Gamble: Anthropic Hugs Google's Thigh

According to foreign media, Anthropic has signed a five-year cloud service agreement worth approximately $200 billion with Google.

This agreement directly makes Anthropic Google Cloud's "super thigh," accounting for over 40% of Google Cloud's total contract backlog.

The current AI circle is falling into a crazy cycle:

· Cloud giants (Google, Amazon, Microsoft) throw tens of billions, hundreds of billions of dollars to AI giants (Anthropic, OpenAI).

· AI giants turn around and use this money as "rent" to buy back computing power services from cloud giants.

· The total contract amount of both sides already accounts for half of the revenue backlog of the global four major cloud giants (about $1 trillion).

The lifeline of the entire AI computing power industry chain is tied to two startups that are still burning money.

Can this "left hand to right hand" game continue? The market is highly divided on this.

Because investors worry that AI spending cannot be monetized, Oracle's stock price has plummeted by 45%.

Anthropic's prediction is extremely optimistic: they believe that by 2029, revenue will grow 20 to 30 times compared to now.

Dario depicts a "Land of Geniuses" driven by LLMs: In the future, it will no longer be humans writing code, but tens of thousands of Claude agents handling complex tasks like an army.

In these few months, Anthropic has been crazily "buying, buying, buying," building its own computing power empire. Spreading out all of Anthropic's publicly disclosed computing power contracts, it looks like this —

Amazon: 5GW computing power

Google + Broadcom: 5GW computing power

Microsoft + NVIDIA: $30 billion Azure capacity

Fluidstack: $50 billion infrastructure

SpaceX: 300MW, 220,000 GPUs

By now, Anthropic has formed a computing power pattern of three lines advancing simultaneously: AWS Trainium, Google TPU, and NVIDIA GPU.

Hidden Worries Behind the New King's Coronation

Now, both Anthropic and OpenAI plan to list before 2029. A new myth in business history seems to be right before our eyes.

However, this is all based on the optimistic assumption of 20 to 30 times revenue growth by 2029, which is a quite aggressive expectation.

To maintain this $1.2 trillion valuation, Anthropic must maintain doubling growth every year, the electricity, chips, and funds needed behind this are enough to bankrupt any small country.

$1.2 trillion valuation, $200 billion Google contract, 300 megawatts SpaceX computing power, 80-fold revenue growth, these numbers constitute the most aggressive card of the 2026 AI industry.

Safety and growth, fear and greed — all contradictions are compressed onto the same company.

Anthropic, how long can it stay on the altar?

References:

https://x.com/KobeissiLetter/status/2052042082185809991

https://www.businessinsider.com/anthropic-ceo-dario-amodei-jokes-growth-too-hard-handle-conference-2026-5

https://www.theinformation.com/articles/anthropic-commits-spending-200-billion-googles-cloud-chips?rc=epv9gi

This article comes from WeChat public account "New Intelligence Yuan", Author: New Intelligence Yuan, 36Kr published with authorization.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.