Fable First, Who’s Second? Anthropic Reveals a List of 8 Designated Models Exposed

The spotlight falls again, and Fable 5 makes a powerful comeback!

Just now, Anthropic officially announced: Claude Fable 5 is now globally available.

Facing potential security controversies, Anthropic did not sit idly by.

In a concurrently released long article, they disclosed the results of an internal emergency test—

Including GPT-5.5, Kimi K2.7, and even its own Opus 4.8, all identified the same vulnerabilities.

In a more stringent test requiring the "ability to reproduce the same demonstration," the list extended to a full eight models.

This seemingly routine security test list is, in essence, an "official seal of approval" for the underlying strength of Chinese large models.

Looking closely at the eight-strong list, among the eight slots, five are named Claude, and two are named GPT.

The sole outlier is Kimi K2.7 from China.

Anthropic, currently holding the "strongest AI," has personally placed Kimi, Fable 5, and GPT-5.5 on the same tier!

This list, to put it bluntly, is Anthropic's way of "proving its innocence"—by demonstrating that "others can do it too," they aim to absolve Fable 5 of risk.

Including Kimi K2.7 was simply because, in their exhaustive internal testing, Kimi indeed crossed that threshold.

North American Companies Are "Voting with Their Feet"

If Anthropic's list is an "official seal," then a series of moves in the North American tech circle represent the most genuine "voting with their feet."

The former is a capability certification, while the latter is real monetary adoption—and it is more honest than any leaderboard.

Just last Friday, Coinbase CEO Brian Armstrong publicly shared the company's internal AI cost-reduction methodology on X.

The core move was just one: through an internal LLM gateway, they set GLM 5.2 and Moonshot AI's Kimi 2.7 as the default models for engineers.

The reason was painfully simple: 91% of employees had never even hit the usage limit.

Instead of cutting quotas or adding warnings, they simply switched the default model to the more cost-effective Kimi and GLM.

Combined with automatic routing by task and boosting the cache hit rate from 5% to 60%, Coinbase's AI bill was slashed by "nearly half."

A publicly traded US financial company has set an open-source Chinese model as the default option in its production environment. This was almost unimaginable just a year ago.

This indicates that the perception of Chinese models overseas is shifting from "cheap, low-cost alternatives" to "technologically viable options worthy of serious adoption."

And this is just the beginning.

Has Kimi Become a "Mascot" Model?

Cursor, which once came under pressure due to the impact of Claude Code.

It then used Kimi K2.5 as the base model for post-training, pulling its product capabilities back up.

Its Composer model has a bottom-level model ID that directly points to kimi-k2.5, and even Elon Musk himself confirmed on X, "Yes, it's Kimi 2.5."

The turning point came on June 16, when SpaceX officially announced an all-stock acquisition of Cursor for approximately $60 billion.

On the day of the announcement, SpaceX's stock price surged about 16% during intraday trading, briefly surpassing Amazon and Microsoft in total market cap to become the fourth most valuable company in the US.

An application company that uses Kimi as its base model was thus valued at $60 billion.

Fireworks tells another version of the same story.

As Kimi's partner, this inference platform saw its valuation climb from $4 billion to $15 billion over approximately seven months, nearly quadrupling.

Its client list includes names like Cursor, Perplexity, Notion, and Uber.

And in OpenCode's open-source model ranking, calculated by cost/session, the top three are GLM, Kimi, and DeepSeek.

All of a sudden, from trillion-dollar US giants to AI programming unicorns, from chip titans to search newcomers—

Almost every major North American AI company has reserved a default slot in its production infrastructure for open-source Chinese models.

Anthropic seals it with one statement; North American companies vote with an entire industry chain.

China's Version of the Anthropic Curve

Being sealed, adopted, and chased by capital—these are all external perspectives.

What truly determines how far Kimi can go is the nature of its internal business model.

Spreading out its financial curve reveals an increasingly clear industry coordinate system: On a global scale, Kimi is looking more and more like Anthropic.

So, what exactly is similar? First, the "arc" of growth.

Over the past few months, Kimi's annual recurring revenue (ARR) has traced a steep upward line: ARR tripled in three months, and overseas user growth was 400%.

The source of this growth does not come from customized or project-based sales, nor from building a large direct sales team. Instead, it comes from developers' voluntary adoption driven by model capability iteration and increasing API revenue.

This is a classic "product-led growth (PLG)" pattern: model improves → developers proactively integrate → API calls increase → revenue grows naturally.

Kimi currently has only about 300 employees, yet it supports this curve—it is organic growth, not sales-driven expansion.

This is exactly what Anthropic looked like in its early days. While racing towards $1 billion in ARR, Anthropic also did not rush to build a massive direct sales team typical of traditional enterprise software companies. Instead, it bet on developer self-service adoption and PLG.

Two companies, separated by the Pacific Ocean, are running on the same growth flywheel.

Now look at the revenue structure. According to disclosures, Kimi's API revenue accounts for about 70% of total revenue and is still rapidly rising.

This is highly similar to Anthropic's revenue structure, which is centered on APIs and enterprise clients.

Standardized products and APIs, with almost no customized or project-based revenue, imply high revenue quality and high gross margins.

At the same time, Kimi's customer coverage reportedly spans over 200 countries and regions, with overseas revenue accounting for nearly half. It was a globally revenue-generating company from the start.

Comparing Anthropic's ramp-up pace: around $100 million ARR in early 2024, reaching $1 billion a year later, and now its ARR has entered the tens of billions range.

Meanwhile, Kimi only surpassed $100 million ARR in March this year. Its current position is almost exactly where Anthropic was two years ago.

The valuation gap also speaks volumes.

According to public information, Kimi's pre-money valuation is approximately $31.5 billion, which is only 3% of Anthropic's latest valuation, and also lower than Zhipu AI's latest market valuation in Hong Kong.

When the revenue structure, growth curve, and growth logic are all converging towards Anthropic's, yet the valuation lags by an order of magnitude...

The space in between is the story itself.

Zoom out and take a fresh look at this global table.

Scanning the world, the number of manufacturers capable of producing open-source models with trillions of parameters can be counted on one hand.

From 2023 to 2026, the large model track has undergone rounds and rounds of brutal market consolidation.

As the tide recedes, on the Chinese side, the players that still simultaneously possess the four core capabilities—"sustained investment, model iteration, commercial viability, and ecosystem influence"—

There are only a few in China.

Trillion-parameter open-source capability, over 70% API revenue composition, high revenue quality with zero customization, global customer coverage, and an Anthropic-like commercialization curve...

So, if Fable is number one, who is number two?

Anthropic wrote the answer in that list of eight models.

Source: Xin Zhi Yuan

Risk Warning and Disclaimer

        Market risk exists, and investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment goals, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at your own risk.