~/LLM/ant-group-introduces-ling-3-0-flash-fin-finance-model-set-to

Ant Group Introduces Ling-3.0-flash-Fin Finance Model, Set to Open-Source Next Week

Ant Group has launched Ling-3.0-flash-Fin, a finance-specialized Mixture of Experts (MoE) language model with 124 billion total parameters and 5.1 billion active parameters. The model's weights will be open-sourced next week, and its API is currently available for a one-month free trial on OpenRouter. This model significantly enhances the processing of complex financial documents like annual reports and spreadsheets while maintaining deployment efficiency. Additionally, the release of the accompanying FinFIRST benchmark will help standardize the evaluation of financial AI agents. The model was optimized through financial corpus pre-training, domain post-training, and tool-use optimization, showing strong performance on benchmarks like SpreadsheetBench and τ³-Banking. It achieves a score increase from 38 to 41 on the AA Intelligence Index v4.1.1, demonstrating concurrent improvements in both general and domain-specific capabilities.

## BACKGROUND

A Mixture of Experts (MoE) architecture scales model capacity by routing inputs to specific subnetworks (experts) rather than activating all parameters for every token, which improves computational efficiency. Evaluating financial AI requires specialized benchmarks like SpreadsheetBench to test complex reasoning over structured data like spreadsheets.

## REFERENCES

## KEYWORDS

#LLM#Finance AI#Open Source#Mixture of Experts

$ subscribe --daily

Ant Group Introduces Ling-3.0-flash-Fin Finance Model, Set to Open-Source Next Week | Daily News