Ling-3.0-flash-Fin - Ant Ling's first finance-enhanced model

What is Ling-3.0-flash-Fin

Ling-3.0-flash-Fin is the first finance-enhanced model launched by Ant Ling, based on theLing-3.0-flasharchitecture (124B total parameters / 5.1B activated parameters), oriented toward real financial workflows. Through financial corpus pretraining and tool optimization, the model strengthens information retrieval, research reasoning, valuation modeling, and research report writing capabilities, and can handle complex content such as annual reports and financial workbooks. It performs outstandingly on financial agent benchmarks such as FinFIRST, with strong competitiveness among models of the same size.

Main features of Ling-3.0-flash-Fin

• Information retrieval: Prioritizes locating official, primary, and highly credible financial materials, ensuring authoritative sources, accurate timing, consistent definitions, and traceable conclusions.
• Research reasoning: Synthesizes cross-document, multi-definition information, building a complete and rigorous evidence chain through multi-step analysis and computation to form reviewable research judgments.
• Valuation modeling: Understands complex financial articulation relationships, automatically reads and updates Excel financial models containing thousands of formulas, and handles actual value overwrites, cross-sheet dependencies, and chart synchronization.
• Research report writing: Identifies core propositions from high-density materials, organizes data, analysis, and charts to form a professional "fact—explanation—judgment" narrative and visual deliverables.

Technical principles of Ling-3.0-flash-Fin

• MoE architecture inheritance: Continues the model architecture and long-context capabilities of Ling-3.0-flash, maintaining a sparse activation configuration of 124B total parameters / 5.1B activated parameters, balancing performance and deployment efficiency.
• Continual pretraining on financial corpora: On top of the general foundation, introduces large-scale, high-quality financial domain data for continual training, strengthening understanding of professional content such as annual reports, financial workbooks, and research materials.
• Domain post-training alignment: Jointly with leading domestic financial institutions and industry experts, deeply participates from task definition, data systems, and training methods to evaluation standards, aligning the model with the professional methods, analytical logic, and compliance requirements of the financial industry.
• Tool use optimization: Optimizes tool-calling capabilities for financial workflows, collaboratively using tools such as Web Search, Visit, and Python under the ReAct agent framework to achieve an end-to-end closed loop of retrieval, computation, modeling, and document editing.
• Coordinated improvement of general and professional capabilities: While enhancing financial domain capabilities, training optimization raises the AA Intelligence Index from 38 points to 41 points, achieving a balance between strengthening professional capabilities and consolidating general capabilities.

How to use Ling-3.0-flash-Fin

Core advantages of Ling-3.0-flash-Fin

• Efficient MoE architecture: 124B total parameters / 5.1B activated parameters, achieving financial task performance close to large-scale flagship models with a smaller activation amount, balancing performance and deployment cost.
• Real workflow closed loop: Integrates information retrieval, research reasoning, valuation modeling, and research report writing into an end-to-end task chain, directly delivering editable and reviewable research outputs.
• Authoritative, traceable sources: Excels at prioritizing official, primary, and highly credible sources, scoring 84.65 on the source selection dimension in the FinFIRST evaluation, ensuring accurate data time points, consistent definitions, and traceable conclusions.
• Synergistic improvement in professional and general capabilities: While financial capabilities are enhanced, the AA Intelligence Index rises from 38 to 41, avoiding an "unbalanced" profile.
• Deep industry co-construction: Ant Group, together with leading financial institutions such as CICC and more than 50 financial professionals, participates in model development and evaluation benchmark design, aligning with real business needs.

Comparison of Ling-3.0-flash-Fin with similar competing products

Application scenarios of Ling-3.0-flash-Fin

• Investment research information retrieval and traceability: Analysts input research questions, and the model automatically retrieves authoritative sources, distinguishes data definitions, traces key conclusions back to original materials, and generates a cited research draft.
• Automatic updates of financial models: During earnings season, the model reads a company's latest quarterly report and existing Excel valuation model, automatically replaces forecast values with actual values, synchronizes cross-sheet dependencies, checks circular references, and updates Summary charts.
• Analysis of complex transaction structures: In investment banking scenarios such as LBOs and M&A, the model traces formula dependencies along operating expenses → EBITDA → free cash flow → debt repayment → IRR, completes multi-scenario return calculations and sensitivity analysis, and outputs an editable workbook.
• Automatic generation of professional research reports: Based on scattered research materials, the model identifies core theses, builds a "fact—explanation—judgment—chart" evidence chain, and generates a draft research report containing visual charts, narrative structure, and professional expression.