AI company Anthropic caused a stir in the market this summer when it introduced Claude Fable and Mythos 5, after which the US government restricted access to the models because of their highly advanced cybersecurity capabilities. With the situation now resolved to some extent, the company has introduced Claude Fable 5.1 – an updated model that promises better results on complex tasks, lower usage costs, and greater efficiency. Let’s take a look at the most important changes in the new version.

Let’s Briefly Revisit What Happened Before Claude Fable 5.1

Unsurprisingly, this model is attracting attention not only because of its technical improvements but also because of its history.

Just three days after the introduction of Claude Fable 5 and Mythos 5, the US government ordered access to the models to be restricted, citing national security concerns. Anthropic then temporarily suspended their use for all customers. Discussions focused on the potential circumvention of the model’s safeguards, although the company stated that similar results could also be achieved with other publicly available AI tools.

Following the lifting of export restrictions, Fable 5 became available worldwide again on July 1. However, the strengthened safeguards also had a downside – they could incorrectly block legitimate requests. When introducing Claude Fable 5.1, the company therefore also focused on making security control less disruptive to everyday work.

What’s New in Claude Fable 5.1

The new model version introduced by Anthropic includes several improvements and changes, which we can divide into the following categories:

Performance

According to the results presented by Anthropic, this version of Claude Fable has better coding skills and seeks specific causes rather than shortcuts. It is also more capable of solving complex tasks. This means the model will handle work requiring multiple stages of analysis or specific decisions more successfully.

Efficiency

The results show that Claude Fable 5.1 achieves the same or better results than the previous version at low or medium effort levels, all at a lower cost.

Low and Medium indicate how much effort the model devotes to solving a task. This means Fable 5.1 can match the previous version’s capabilities without using the highest effort level, making it easier for users to balance the quality of results with usage costs.

Security

Previous events have already shown that cybersecurity is particularly important as modern models become increasingly advanced. Anthropic highlights that this time, the model will interrupt legitimate work due to unfounded suspicions approximately 60% less often. However, restrictions on creating certain tools will remain, and Claude Fable will also identify software vulnerabilities.

Data privacy

Another feature of Claude Fable 5.1 is the Enterprise Frontier Safeguards (EFS) system for businesses. It will be introduced later in the autumn and will provide data storage within customer-controlled infrastructure. Until then, some customers will be able to use a zero data retention option.

Pricing

According to Anthropic’s calculations, using Fable 5.1 for typical tasks should cost approximately 25% less than using the previous version, while savings on work requiring many autonomous actions by the model could reach around 45%.

This reduction is largely due to the lower cost of reading previously processed and stored information from the cache. This means that when working with the same information across several stages, reusing it becomes cheaper. Meanwhile, base prices will remain unchanged: one million input tokens will cost $10, while one million output tokens will cost $50.

What Can Claude Fable 5.1 Do in Practice?

The company also evaluated the model on the specific tasks. For example, in tests conducted by investment firm Millennium, the systems identified the cause of a rare issue that caused internal systems to freeze.

In scientific research, Claude Fable 5.1 trained a neural network and created a more detailed elevation on map of one-third of Venus’s surface using old images from one of NASA’s radars.

It is important to note that Claude Mythos 5.1 was introduced alongside the model. It is based on the same underlying model but has more restrictions and more limited availability.

Final Thoughts

After the turbulent history of the Fable 5 launch, the new version brings attention back to the model itself and its benefits for users. Anthropic aims to make advanced capabilities easier to apply in practice – from complex coding tasks to scientific research. So, Claude Fable 5.1 could become a more attractive option for those whose use of the previous version was limited by costs or overly frequent request blocking.

If you are interested in this topic, we suggest you check our articles:

• Fable 5.1 vs GPT-6 Astra: Which Model Executes Tasks Best?
  
• Should Developers Use Fable or Mythos When Using Claude?
  
• Why the US Government Changed Its Stance on Claude Fable 5 Access
  

Source: Anthropic