Anthropic Unveils New AI Models Fable and Mythos Ahead of IPO

Anthropic has unveiled two powerful AI models as the company prepares for a potential IPO on Wall Street. With Fable 5.2 and Mythos 5.2, the creator of chatbot Claude wants to further increase competition with OpenAI. The new models are faster and can perform more complicated tasks, but also raise concerns about the safety of artificial intelligence.

Anthropic releases Fable 5.2 on Monday, November 9. According to the company, the new AI model can carry out more complicated programming tasks, control robots and work independently on projects for a long time.

Remarkably, Fable can coordinate more than 1,000 AI agents at a time. These are digital assistants who independently perform tasks and collaborate on larger assignments.

In addition to improved performance, Anthropic also reduces costs. The price of processing one million imported tokens drops from $10 to $8. For a million generated tokens, the price goes from $50 to $40. This means a 20% decrease in both cases. Tokens are small pieces of text that an AI model processes.

At the same time, Anthropic introduces Mythos 5.2. This model uses the same technology, but has fewer limitations for sensitive applications, such as software vulnerability detection and life sciences research.

Due to the potential risks, Mythos is not available to everyone. Only selected organizations will be granted access through a dedicated verification program.

The new AI models appear at an important time for Anthropic. The company is preparing for an IPO on Wall Street, which according to Bloomberg may take place later this month.

With the launch, Anthropic wants to strengthen its position in the competitive battle with OpenAI. Both companies are investing billions in the development of increasingly powerful AI systems.

For Anthropic, it is especially important to attract more business customers. The new models focus, among other things, on companies that use AI for software development and the automation of complex work.

Price plays an important role in this. Figures from payment platform Ramp previously showed that Anthropic's most expensive AI models accounted for only eleven percent of total spending on its models.

By making Fable cheaper, Anthropic hopes more companies will choose its most powerful technology. This can be important for further growth in turnover and convincing investors in the run-up to the IPO.

The introduction of the new models also creates discussion about the safety of artificial intelligence. Mythos is particularly sensitive because of its possibilities in the field of cybersecurity.

The first version of Mythos, released in April, was able to detect serious vulnerabilities in software. At the same time, there was a risk that the technology could also be used to exploit those vulnerabilities.

The US government therefore restricted access to the most powerful models earlier this year. Later, the restrictions for certain organizations were relaxed.

Remarkably, in September, Anthropic CEO Dario Amodei himself called for the development of the most powerful AI systems to be slowed down. Shortly before, a researcher at the company had left because of concerns about the potential dangers of AI.

According to Anthropic, Fable 5.2 and Mythos 5.2 have therefore been tested more extensively than previous versions. Among other things, it was investigated whether the systems can escape from closed digital test environments.

Dianne Penn, responsible for research products at Anthropic, says the new models were less likely to exhibit undesirable behavior during these tests. The US government and independent researchers have also assessed the technology in advance.

At the same time, political pressure is increasing. After the Democratic victory in the US midterm elections, stricter rules may follow for AI security and the construction of data centers.