Kolleno
AI-driven accounts receivable automation to accelerate cash flow and optimize collection workflows.
The first 50-billion parameter LLM purpose-built for the global financial industry.
BloombergGPT is a proprietary, 50-billion parameter large language model (LLM) specifically engineered to master the nuances of financial data and terminology. Developed by Bloomberg's AI researchers, the model is trained on 'FinPile'—a massive 363-billion token dataset consisting of financial documents, news, and filings collected over 40 years, supplemented by general-purpose text. By 2026, BloombergGPT has been fully integrated into the Bloomberg Professional Services ecosystem, serving as the core engine for sentiment analysis, financial entity recognition, and automated market research. Unlike general LLMs, BloombergGPT excels at domain-specific tasks such as interpreting ticker symbols, understanding fiscal quarters, and parsing regulatory language with high precision. Its architecture is a decoder-only transformer that utilizes a hybrid training objective to maintain general linguistic proficiency while dominating financial benchmarks. For institutional investors and analysts, it provides a unique competitive edge by reducing the time required to synthesize vast quantities of market data into actionable insights, operating within the high-security, high-reliability environment of the Bloomberg Terminal.
Trained on a 363-billion token financial-specific dataset including 40 years of market archives.
AI-driven accounts receivable automation to accelerate cash flow and optimize collection workflows.
The Intelligence Layer for Global Financial and Professional Services Data.
Institutional-grade AI-driven investment strategies and quantitative risk management through the Marquee platform.
Precision algorithmic engine for institutional-grade price-gap detection and real-time market dislocation analysis.
Verified feedback from the global deployment network.
Post queries, share implementation strategies, and help other users.
Specialized tokenizer that recognizes and processes financial symbols and shorthand as primary entities.
Architectural tuning to prioritize the recency and relevance of market news over historical data.
Combines masked language modeling with financial-specific sequence prediction.
Deep learning layers optimized for identifying financial entities (PE firms, CEO names, subsidiaries) without specific fine-tuning.
Automated extraction of ESG-relevant metrics from unstructured corporate reports.
Operates within Bloomberg's private cloud with strict data residency and audit trails.
Analysts spend hours listening to calls and reading transcripts to find key guidance changes.
Registry Updated:2/7/2026
Generate a 3-bullet summary for the trading desk.
Social media and news noise can hide genuine market-moving sentiment shifts.
Identifying subtle but critical changes in 10-K risk factors year-over-year.