An open-weight AI model brings financial research, valuation modelling and reporting into one workflow, combining MoE efficiency with traceable information retrieval and spreadsheet automation.
Ant Group has open-sourced Ling-3.0-flash-Fin, an AI model designed for financial research and analysis workflows where reliable information, accurate calculations and traceable outputs are critical. Unlike models focused primarily on conversational question answering, the new model is designed to work with dynamic financial information, accounting requirements and professional research processes.
The model uses a Mixture-of-Experts (MoE) architecture with 124 billion total parameters, while activating only 5.1 billion parameters per token. This approach aims to combine the knowledge capacity associated with larger models with lower inference costs and more efficient deployment. The source reports competitive performance across financial benchmarks including FinFIRST, FinSearchComp Verified, FinCRAFT, FinanceAgent, APEX-Agents, SpreadsheetBench and τ³-Banking.
Four capabilities shape the workflow
Ling-3.0-flash-Fin focuses on four areas relevant to professional financial work.Information retrieval prioritises authoritative sources to improve data consistency and maintain end-to-end traceability. Research reasoning combines heterogeneous information from multiple sources to build logical, verifiable evidence chains. Valuation modelling enables the model to understand linked Excel financial models and automate updates while keeping spreadsheets editable. Report generation combines research facts, calculations and charts into structured financial outputs.
This combination makes the model particularly relevant to investment research, where an AI system needs to move beyond generating text and instead connect source data, reasoning, calculations and presentation. The model is available through OpenRouter and Vercel, while its open weights are available through Hugging Face and ModelScope. Developers can deploy it privately and connect it with search, Python, databases and spreadsheets for customised financial workflows.
Alongside the model, Ant Group has also open-sourced FinFIRST, a benchmark for financial search agents. Developed with professional support from China International Capital Corporation, FinFIRST V1 contains 123 expert-authored tasks, 701 atomic criteria and 12,300 rubric points. Rather than evaluating only final answers, it assesses the broader research process, including data consistency and traceability. The release is part of the broader Ling 3.0 family, which also includes models targeting production agents, local deployment and multimodal workloads.
