Anthropic releases new powerful models ahead of Wall Street debut

Anthropic announced on Monday (9) the launch of two new iterations of its artificial intelligence models: Claude Fable 5.2 and Claude Mythos 5.2. These tools incorporate notable improvements in areas such as reasoning, programming, and visual interpretation, although their launch occurs during an intense debate about AI's ability to drive its own development.

According to the company, the newly released models demonstrate greater clarity in communication and offer more agile responses. Specifically, Fable 5.2 saw its entry fee decrease from US$ 10 to US$ 8 per million tokens, representing a 20% reduction (equivalent to about R$ 43). Furthermore, Anthropic reports that this model's responses became 30% faster across various complexity levels.

Access to Mythos 5.2 will be limited to organizations validated by the company itself, including those involved in cybersecurity focused programs. On the other hand, Fable 5.2 is already accessible through the Anthropic API and cloud computing platforms.

This launch follows a report issued by the AI safety organization, METR, just a few days prior. This study indicated that Anthropic's developed models are progressing almost twice as fast as before, an advancement attributed partly to the use of AI in the development cycle itself.

The analysis, conducted over eleven weeks and published on November 5, calculated that the performance of the company's models in well-defined tasks advanced 1.9 times faster since October 2025 compared to the preceding period. This finding reignited conversations about the concept of recursive self-improvement (RSI), which describes the potential for AI systems to assist in improving future versions of themselves, thereby accelerating consecutive technological generations.

However, METR made important caveats to its study, classifying the estimate as highly uncertain; the actual acceleration rate could vary between 1.5 and 2.5 times the historical pace. The study also highlighted that AI does not yet have the capacity to generate highly relevant contributions fully autonomously in the crucial phases of model training.

Additionally, Anthropic informed that Fable 5.2 was used in scientific research before its public release. Results credited to this model include solving two pending problems in theoretical physics and formalizing a proof for Fermat's Last Theorem. The company also mentioned that its models helped engineers finalize software migrations considered impractical, conceive three functional electronic devices from scratch, and advance in robotics-related evaluations.

In another disclosure made on Monday, the startup Physical Superintelligence presented a new demonstration regarding the Aaronson-Ambainis conjecture, a topic relevant to defining the theoretical limits of quantum computation. This company claimed to have achieved this result using an undisclosed Anthropic model, although the developer did not confirm whether the specific tool was Fable 5.2.

Artificial intelligence (AI) may be driving its own progress. An independent examination conducted by the non-profit organization Model Evaluation and Threat Research (METR) indicated that models developed by Anthropic are progressing almost twice as fast as before. This is partly due to the company's own use of AI systems to create more powerful versions of its technology.

The study, published this Thursday (5th), calculates that the advancement in the capabilities of Anthropic's models in specific tasks increased at a rate of 1.9 times since October 2025, compared to the previous pace.

According to METR, approximately 80% of this current acceleration is linked to the application of AI in the company's research and development activities. However, the estimate has a high degree of uncertainty.

The evaluation was carried out over an 11-week period, covering August to October 2026. Researchers had access to Anthropic's internal data, interviewed employees, and analyzed information on the performance of models that have not yet been made public.

The organization also found that the increase in computational capacity used for training does not justify the observed acceleration. The study establishes a crucial distinction: most of the progress still derives from the assistance provided by AI to experts, not from the autonomous execution of the entire development cycle.

METR stated that it found no evidence that current systems can offer highly relevant and fully independent contributions in the most critical phases of model training. Despite this, researchers detected signs that such participation may become more prominent in the future.

The advance noted by METR deserves attention because it is connected to the concept of recursive self-improvement, known by the acronym RSI. This idea posits that AI systems can increasingly contribute to the generation of new model versions. By assisting in the development of their successors with more competent systems, the process tends to gradually accelerate.

Although this scenario has not yet been achieved in a completely autonomous manner, it raises important questions about the speed of technological evolution. One concern highlighted by METR is the potential for this acceleration to become cumulative, where more advanced models assist in developing even more capable systems, which in turn would intensify the process.

The organization also did not identify clear impediments related to computing power, data availability, or experiments that would certainly block new advances in the next twelve months. However, the researchers emphasized that this analysis is also subject to uncertainties.

Despite the findings, METR recognizes significant limitations. The 1.9-fold acceleration was calculated based on performance indicators in well-defined tasks, which can be assessed through automated tests or language models. This does not imply that all aspects of AI are evolving at the same pace.

Additionally, Anthropic had the prerogative to remove confidential information from the report. According to METR, much of the evidence considered most pertinent to its conclusions was omitted, which complicates an independent assessment of all details.

The organization also highlights the uncertainty about what portion of the acceleration can be directly attributed to the use of AI, especially regarding the progress recorded at the end of 2025.

The results emerge at a time when Dario Amodei, CEO of Anthropic, advocates for prudence in the development of increasingly sophisticated systems. In September, the executive called on the sector to slow down the pace of technological capability advancement so that safeguards could keep up with the pace of improvements.

The study itself points out that understanding the speed of this evolution is vital for risk assessment. If systems begin to collaborate in a progressively more significant way in the development of their replacements, monitoring and control work may also become more complex.

Anthropic was created by a group of former OpenAI employees in 2021 who decided to found a new company focused on the safe development of artificial intelligence. The company was named Anthropic, meaning 'related to humanity.' Its debut on Wall Street is expected soon and could become one of the largest initial public offerings in history.

However, this move raises a fundamental question: is it possible for a company to put humanity first while meeting the demands of Wall Street? The IPO represents a significant event for the AI boom and will draw attention to how investors view the developer of the Claude chatbot. This entry into the stock market marks a new chapter for Anthropic, which will test the company's ability to balance its mission of safe AI development with the pressures associated with being a publicly traded company.

Anthropic uses a special ownership structure designed to help the company adhere to its original goals, balance profit growth with its mission, and prevent acquisition by investors. This structure includes a board of trustees that holds majority control of the board of directors, as well as a special class of shares guaranteeing that the seven founders hold the majority of votes among general shareholders.

Due to its mission of ensuring AI safety and its unconventional corporate structure, Anthropic's debut on Wall Street can be seen as a test for cutting-edge AI labs and how they are perceived by investors. Anthropic is registered as a Public Benefit Corporation (PBC). Such companies are legally obligated to balance shareholder interests with the stated company mission, distinguishing them from traditional companies primarily focused on maximizing shareholder value.

Gil Luria, Head of Technology Research at the financial firm D.A. Davidson, noted: 'Anthropic's governance structure is very unique. They write it down. They are very open about it.' In its S-1 filing, Anthropic described its 'three-part goals' as a PBC as follows: 'the financial interest of our shareholders, the best interest of those substantially affected by our activities, and the Company Goal.'

Only a small fraction of publicly traded companies are PBCs. Patagonia, which is private, and the optical retailer Warby Parker (WRBY) are also PBCs. PBCs can independently decide what to do with corporate profits and invest as much as they wish in safety measures, with little opportunity for investors to influence this.

Anthropic's competitor, OpenAI, became a PBC in 2025. Luria added: 'Anthropic explicitly states that they believe there will be times when they will choose not to maximize shareholder value based on judgments made by the board of directors and management team.' He continued: 'Company growth and profitability are part of the goal, so this does not contradict the goal. They just say there may be times when it is not the goal, or they pursue other goals, even at the expense of growth and profitability.'

The seven founders of Anthropic collectively own 50.1% of the company's votes, meaning ordinary shareholders will have no influence over its operations. This is a variation of the trend in large technology companies in recent years, where structures have provided for retaining control by the majority of founders, similar to Mark Zuckerberg at Meta and Elon Musk at SpaceX.

It is not only the founders who hold the majority of votes. In 2023, Anthropic established a Long-Term Benefit Trust (LTBT), composed of five experts in relevant fields who have no financial stake in the company's shares. However, they possess special non-financial shares with the authority to appoint four directors to the company's seven-member board.

Members of the LTBT include Neal Buddy Shah, CEO of the Clinton Health Access initiative, and Ben Bernanke, former Chairman of the Federal Reserve. There is one vacancy in the LTBT yet to be filled. Anthropic reports in its SEC filing to shareholders: 'The LTBT may have interests that differ from yours, and it may exercise its power in a way you do not agree with, or in a way that may be unfavorable to your interests.'

When the company goes public, there may be increased pressure to increase profits and achieve strong results for shareholders. However, Anthropic and CEO Dario Amodei have also discussed the risk that AI could cause immense harm to humanity if it falls into the wrong hands. Anthropic's IPO filing contains over 90 pages of risk disclosures, ranging from how AI could be misused by malicious actors to the possibility of existential risk to humanity.

Jay Ritter, Professor Emeritus at the University of Florida and an IPO expert, stated: 'Although the LTBT has a lot of power, the founders also have a lot of power. But, you know, in both cases, public shareholders have no power.'

If growing concerns about AI arise, the founders and the LTBT can make decisions for the company that affect stock value without regard for investor interests. Theoretically, this could help the company adhere to its mission—but it remains unknown how the structure will function until some event forces the company to act.

Anthropic's ownership structure will matter less if profits are high and the company adheres to its mission of developing safe AI. However, it will be tested if revenues fall below expectations, security issues arise, or disagreements emerge between the board of directors, founders, or trustees, as noted by Luria of D.A. Davidson.

Jay Goldberg, Senior Analyst at Seaport Research Partners, wrote in a note: 'The strangest aspect of Anthropic's filing is its corporate governance structure. We believe it is the largest PBC, and the full implications of this structure will take time to unfold.'

Goldberg added: 'We believe this structure gives the LTBT and the founders effective permanent control over the company.'