---
title: "New versions of Anthropic's models unveiled"
url: https://noti.group/new-versions-of-anthropic-s-models-unveiled/
language: en
publisher: "Noti Group"
section: "Technology"
published: 2026-11-09T19:24:51.000Z
updated: 2026-11-09T20:28:53.132Z
id: f035040a-24eb-4187-bdd6-0a82fb5fbe9c
source: "TechCrunch https://techcrunch.com/2026/11/09/anthropic-releases-new-mythos-and-fable-models-amid-rsi-concerns/"
attribution: "Link to https://noti.group/new-versions-of-anthropic-s-models-unveiled/ and name Noti Group when you quote or summarize this story."
---

# New versions of Anthropic's models unveiled

Anthropic has released updated versions of its Mythos 5.2 and Fable 5.2 models, designed to improve performance in coding, reasoning, and vision tasks.

Anthropic has unveiled new versions of its advanced models, Mythos 5.2 and Fable 5.2, which are designed to improve performance in coding, reasoning, and vision tasks.

The company claims that these upgraded models can communicate more effectively with users, a notable improvement over their predecessors. The enhanced capabilities of the new models aim to address some of the limitations of previous versions.

Anthropic has also taken steps to make its Fable 5.2 model more competitive in the market by reducing its price from $10 to $8 per million input tokens. This move puts Fable 5.2 at a lower cost than its main competitor, GPT-6 Astra, and is likely to appeal to customers looking for a more affordable option.

The company has also announced that users of Fable 5.2 can expect responses from the model to be delivered 30% faster across different levels of effort, without any compromise on quality or accuracy.

Faster inference serving is now available for double the price, providing an additional layer of speed and efficiency for users who require it.

Anthropic's new Mythos and Fable models have been released, featuring several key updates to their capabilities. The company maintains that the latest version of Mythos does not possess covert action abilities against its user, a claim further detailed in the system card for Mythos 5.2.

According to the system card, Claude Mythos 5.2 exhibits stronger covert capabilities compared to previous models when its reasoning is hidden from view. However, Anthropic believes that low absolute performance on new covert evaluations supports claims of detecting harmful actions in high-stakes situations.

The Fable model has been available in pre-release to select parties for some time and has already made significant contributions to research projects. It has solved two open problems in theoretical physics, formalized Fermat's Last Theorem, and supported numerous other research initiatives.

A separate development saw AI startup Physical Superintelligence announce a new proof of the Aaronson-Ambainis conjecture, a crucial step in setting theoretical limits for quantum computing. This achievement was made possible by running an Anthropic model within PSI's system, although it remains unclear whether Fable 5.2 was the specific model used.

The release of new models by Anthropic has sparked concerns about repetitive sequence inefficiency (RSI) in AI systems. The company's Mythos and Fable models are designed to be more efficient and scalable than their predecessors. However, the use of these models raises questions about whether they can truly mitigate RSI issues.

Anthropic's achievement in creating a model that can perform tasks with minimal repetition is significant, but its reliance on running an Anthropic model within PSI's system has sparked debate. The specific model used, Fable 5.2, has been identified as the one responsible for this breakthrough, although its exact capabilities and limitations remain unclear.

The release of these new models marks a critical juncture in the development of more efficient AI systems, but it also highlights the ongoing challenges associated with RSI. As researchers continue to push the boundaries of what is possible with AI, they must also address the potential risks and limitations of their creations.

---
Source: [TechCrunch](https://techcrunch.com/2026/11/09/anthropic-releases-new-mythos-and-fable-models-amid-rsi-concerns/)  
Published by Noti Group: https://noti.group/new-versions-of-anthropic-s-models-unveiled/