Ling-3.0-flash-Fin: Ant Group Open-Sources a Financial AI Model Built for Research

A financial professional researching a company can spend much of the task collecting documents, checking figures, comparing information and turning the findings into a report. Ling-3.0-flash-Fin is designed to handle those parts of the process, and Ant Group has now released it as an open-source model.

Ant Group announced the release at the 2026 Inclusion·Conference on the Bund in Shanghai on September 9, 2026. The company developed the model with financial institutions and industry experts, incorporating their expertise into its tasks, data systems and evaluation.

What Ling-3.0-flash-Fin is designed to do

The model covers four areas: information retrieval, research reasoning, valuation modelling and report generation.

For someone using AI in financial work, those categories describe a recognisable sequence. Information has to be found first. It then has to be interpreted, used in calculations or models, and turned into an output that another person can review.

Ant Group says financial work also requires trustworthy sources, consistent definitions, accurate calculations and auditable results. Those requirements are part of how the company says it developed and evaluated the model.

The distinction matters because a generated answer is only one part of financial research. The underlying documents, assumptions and calculations can matter just as much when the result is used in a professional setting.

The model is built for long financial tasks

Ling-3.0-flash-Fin uses a Mixture-of-Experts architecture with 124 billion total parameters. It activates 5.1 billion parameters per token.

The model also has a 256K context window. In practical terms, the architecture and context window are intended to let the system work through substantial amounts of material without treating each document or calculation as an isolated question.

Ant Group's description includes spreadsheet work alongside research and valuation modelling. That puts familiar office tasks inside the model's intended use rather than limiting it to conversational answers.

What this means for financial research

The most relevant part of the announcement for a financial professional is the range of tasks the model is intended to cover. A research assignment can involve finding information, reconciling figures, assessing assumptions, building calculations and preparing a written result.

Ling-3.0-flash-Fin is designed to operate across that sequence. Ant Group says it can be connected to search, Python, databases and spreadsheets, which are the kinds of tools that sit around financial research today.

That does not establish how well the model will perform in a particular organisation. It does establish the type of work Ant Group is targeting.

Ant Group is testing the research process too

Ant Group has released FinFIRST alongside the model. The benchmark contains 123 expert-authored tasks, 701 atomic criteria and 12,300 rubric points.

FinFIRST evaluates a financial search agent's research process rather than looking only at its final response. That means the assessment can account for how the agent handles the work that leads to an answer.

The benchmark was developed with support from the CICC investment banking team. Its inclusion gives the release a second component that matters to professionals evaluating AI: a framework for assessing financial research tasks rather than relying only on general-purpose model scores.

The open-source part matters for organisations

Ling-3.0-flash-Fin is available through OpenRouter and Vercel, while Ant Group has released its open weights through Hugging Face and ModelScope.

The model can also be privately deployed and connected to search, Python, databases and spreadsheets. That makes the deployment question relevant for organisations that have their own financial information and existing software systems.

This is part of a wider finance-specific AI effort

Ling-3.0-flash-Fin joins other models in Ant Group's Ling 3.0 series, including Ling-3.0-flash, Ling-3.0-tiny, Ling-3.0-flash-VL and Ling-3.0-flash-Santé.

Ant Group has also worked on AI for a narrower financial task through its finance forecasting model. The new model covers a wider set of activities around financial research and analysis.

For professionals who do this work, the useful detail in the announcement is the list of tasks. Ant Group is putting information retrieval, analysis, valuation, spreadsheets and reporting into one financial AI system, while also releasing a benchmark intended to test how the research gets done.

That gives us a clearer picture of where financial AI is being applied: inside specific pieces of professional work, where the quality of the sources, calculations and resulting analysis still has to be examined by the person using it.

Editor's note

Every piece published on The Bright Minded goes through careful verification, but mistakes can happen. If you spot an error, have additional information, or want to flag anything, write to rosalia@thebrightminded.com.