Ling-3.0-flash-Fin

Ling-3.0-flash-Fin is a finance-specialized large language model developed by InclusionAI, an initiative by Ant Group. It is a domain-specific fine-tune of the Ling-3.0-flash base model, designed to handle high-complexity financial reasoning and investment research tasks. The model was developed in collaboration with leading financial institutions and domain experts to ensure accuracy in professional workflows.

The model utilizes a Mixture-of-Experts (MoE) architecture, featuring 124 billion total parameters with approximately 5.1 billion active parameters per token. This design enables the model to deliver reasoning capabilities comparable to larger dense models while maintaining the inference efficiency and low latency required for real-time financial applications. It supports a native context window of 262,144 tokens (256K), allowing it to process extensive financial filings and multi-document datasets without information loss.

Ling-3.0-flash-Fin is optimized for end-to-end financial research, including source-grounded search that prioritizes authoritative regulatory filings and earnings reports. It excels at multi-document reasoning, reconciling conflicting figures across different reporting periods or materials, and performing complex valuation modeling. Its spreadsheet-native capabilities allow it to understand formulas, cross-sheet dependencies, and scenario analyses, making it suitable for automating financial model updates.

Evaluation results show competitive performance on finance-specific benchmarks such as FinFIRST, FinCRAFT, and SpreadsheetBench. The model supports native function calling, enabling agentic workflows that can interface with external data tools or calculation engines. While highly capable in finance, the model retains the strong general reasoning, mathematics, and coding abilities of the underlying Ling-3.0-flash foundation.