The Hong Kong Telegraph - Cipheras Group - Apex AI Fund Deploys 290-Billion-Parameter Proprietary Large Language Model

Cipheras Group, the investment manager of Apex AI Fund, today announced the full deployment of its proprietary large language model, Apex LLM v2.4, as a core analytical component of the fund's live investment process. The model has reached 290 billion parameters trained on financial and technology-sector data, and is now processing in excess of 2.4 million data points daily with an average signal latency of 94 milliseconds.

The deployment positions Apex AI Fund among the first private investment funds globally to integrate a purpose-built, proprietary large language model into the active daily management of a live equity portfolio.

A Model Built for Investment Intelligence

Apex LLM v2.4 was developed exclusively for financial investment analysis and trained on a proprietary corpus comprising SEC filings, earnings call transcripts, patent application databases, academic AI research literature, GitHub repository activity, corporate job posting data, and structured financial market data. Unlike general-purpose commercial language models, it is purpose-optimised for investment signal generation within the artificial intelligence sector.

The system processes six primary data source categories simultaneously and delivers structured signals to the fund's investment team in near real time. Critically, Apex LLM v2.4 does not make autonomous investment decisions. All signals are reviewed by the human investment team before any portfolio action is taken.

"We did not build this model to replace judgment — we built it to protect it," said Sahasra Deekonda, Chief Risk Officer of Apex AI Fund. "The most dangerous moment in investment management is when you miss something important because there was simply too much to read. Apex LLM v2.4 means we never miss the signal that matters. The decision, however, will always rest with the humans in the room."

Documented Signal Performance

Since deployment, the system has generated multiple actionable signals acted upon by the investment team. Among the most significant: the LLM identified a surge in TSMC capacity booking disclosures six weeks before NVIDIA's Q2 2025 earnings call confirmed the supply ramp, enabling the fund to increase its position ahead of the announcement. The system also flagged governance anomalies in a fund holding within 24 hours of the relevant filing, triggering a position review that materially limited the fund's downside exposure. Additionally, profitability trajectory analysis identified the probability of a major portfolio holding's S&P 500 inclusion eleven weeks before the official announcement.

"Building a large language model trained on financial data is not a marketing exercise for us — it is a competitive necessity," said Dr. Thrisha Ganesh, Co-Founder and Head of Research at Apex AI Fund. "The volume of material that becomes public every single day through filings, transcripts, and research is simply beyond what any human team can absorb at the required speed."

About Apex AI Fund

Apex AI Fund is a private equity fund investing exclusively in publicly traded companies at the forefront of the artificial intelligence revolution, managed by Apex Capital Intelligence LLC. The fund launched in September 2022 with initial positions in NVIDIA, Micron Technology, Palantir Technologies, and a concentrated group of AI infrastructure companies. Since inception, the fund has delivered a cumulative net return of +348%, averaging more than 75% per fiscal year. The fund manages approximately $134 million in assets and accepts investors with a minimum initial investment of $5,000. No management fee, sales fee, or redemption fee is charged. A tiered performance fee applies exclusively to net profits.

www.cipheras.com

Contact

Source article: https://financewire.com/2026/09/07/cipheras-group-apex-ai-fund-deploys-290-billion-parameter-proprietary-large-language-model/

鄭-H.Zhèng--THT-士蔑報