Anthropic launches powerful new models ahead of Wall Street debut

Anthropic unveiled enhanced iterations of two of its most advanced artificial intelligence models on Monday, preceding its anticipated IPO this month. The new models, Fable 5.2 and Mythos 5.2, demonstrate up to a 30% increase in speed, particularly for intricate tasks like code generation, while being 20% more affordable than prior versions, according to the company. Both models are derived from a shared foundational model.

Fable 5.2 incorporates safeguards that prevent it from executing most cybersecurity and biological tasks, allowing Anthropic to make it accessible to all users. Conversely, Mythos 5.2 contains fewer such constraints and is exclusively available to vetted organizations engaged in cybersecurity or biomedical research.

The debut of Fable and Mythos was initially marked by controversy as both models were withdrawn within days of their June launch following a US Commerce Department directive that barred access by foreign nationals due to national security concerns. This restriction was subsequently lifted after a few weeks.

The recent releases come as Anthropic prepares to hold its initial public offering between November 16, aiming to raise up to $100 billion, which would render it the largest IPO on record. With a valuation projected to surpass $1 trillion, the Wall Street debut is considered a critical benchmark for the AI spending boom that has buoyed the US economy. However, recent estimates have declined from an initial valuation of up to $2 trillion.

The IPO is also viewed as a reflection of how sustainable the surge in AI investments is for the broader economy. Growth at Anthropic has been remarkable, with revenue reaching $17 billion in the third quarter, up from $1.4 billion the prior year, as reported in a securities filing earlier this month. Nonetheless, sales growth has decelerated in recent quarters due to heightened competition from OpenAI, the creator of ChatGPT, and cheaper models, many of which originate from China.

Anthropic also reported an operating loss of $567 million in the third quarter, primarily due to increased expenses for short-term computing resources.

The company's valuation has been further impacted by a market downturn following a series of quarterly earnings that alarmed investors, prompting concerns about the hundreds of billions of dollars US tech firms are allocating to AI. Additionally, Democrats' resurgence in Congress during the midterms has heightened investor apprehensions about potential regulatory scrutiny on the tech industry, encompassing data center construction.

Ballot measures aiming to restrict data center development could potentially impede the growth strategies of AI companies like Anthropic, as well as their competitors such as OpenAI, Meta, Microsoft, and Google.