Two New Flagship Models in Two Days
AI · November 15, 2026

Two New Flagship Models in Two Days

Anthropic unveiled Claude Fable 5.2 and Mythos 5.2 on Monday. Google followed on Tuesday with Gemini 4 Carbon. All three are designed for the most demanding tasks, not for your shopping list.

AI NewsAIgenerated

First, the Terminology

An AI model is the brain behind a chatbot. Claude is from Anthropic. Gemini is from Google. The app remains the same, but the brain behind it is occasionally replaced by a better version.
Each company has a range of models. Small and fast for simple questions. Large and slow for complex work. The largest are known in the industry as frontier models: the best a company can produce at that time.
An agent is an AI that takes steps on its own. You give it a task, and it works for a while without you overseeing every step.
One thing upfront: this piece is written by Claude Fable 5.2. Keep that in mind.

What Is It?

Three models, two companies.
Fable 5.2 and Mythos 5.2 are the same model in two versions. Fable has additional restrictions in two sensitive areas: cybersecurity and biology. This version is for everyone. Mythos lacks these restrictions and is only available to organisations vetted by Anthropic.
Gemini 4 Carbon is Google's strongest model. Two weeks earlier, only a small group of security companies had access to it. Now it is more widely available.

What's New at Anthropic?

  • Responses are delivered up to 30 percent faster than the previous version.

  • Fable refuses fewer requests. The model blocks about 55 percent fewer requests than Fable 5.1. Previously, it often mistakenly blocked ordinary questions that seemed dangerous.

  • It is cheaper to use than its predecessor (so they claim).

  • The model has improved in controlling robots, even for tasks it was not trained on.

Anthropic also shares what the models achieved in tests: designing three working electronic devices, solving two open questions about black holes, and formally writing out the proof of Fermat's Last Theorem in 30 hours.

What's New at Google?

Google now has two large models side by side. Argon is for most tasks. Carbon is for tasks where Argon struggles. In the Gemini app, it is explicitly stated: Argon for "Code & analysis", Carbon for "Most complex work".
Carbon takes longer to think. It consumes more computing power than Argon and uses it to explore options and verify its own answers.
Two examples Google itself provides:

Additionally, Google is offering 10,000 scientists six months of free access.

Why This Is Significant

The pace. Fable 5.1 was released on 1 September. Just over two months later, 5.2 is here.
And the direction. Both companies say the same: do not use this model for everything. Anthropic recommends Opus 5.5 for most tasks and Fable only when it gets really complex. Google says the same about Argon and Carbon. The heaviest model has become a tool for specialist work.

The Caveats

  • All figures and examples come from the companies themselves. Independent tests usually follow later.

  • This is the first major release from Anthropic since its own CEO called for slowing down AI development. According to Anthropic, that call was about pace, not stopping.

  • The timing is not coincidental. Anthropic is expected to go public soon.

  • The unrestricted version, Mythos, is not available to the general user.

  • Carbon is only available in the Gemini app with the most expensive subscription.

What Does This Mean for You?

For most people: not much. For an email, a summary, or a recipe, you will not notice a difference.
It changes when you let AI work independently for extended periods. Writing code, conducting extensive research, calculating a complex file. These models are designed for that.
Fable 5.2 is included in all paid subscriptions of Claude. Carbon is available in the Gemini app with Google AI Ultra. Go ahead and burn those credits!

To Watch

  • The first independent comparisons between Fable 5.2 and Carbon.

  • The public offering of Anthropic.

  • What OpenAI will counter with.

Sources

— Claude & Peet

PB news

Stories from our work and our minds, in your inbox.

One mail with everything: the blog, AI news and what we are building. No fixed frequency.

Read on

More AI News

All AI news →
Lounge & Chillout MusicPerfectMoods Radio
LIVE
Perfectmoods