Ling-3.0-flash-Fin Takes Auditable Tool Chains Into Financial Work
- Source
- x.com
- Date
Financial work depends on trustworthy sources, consistent definitions, accurate calculations and auditable outputs. Introducing Ling-3.0-flash-Fin, a finance-enhanced version of Ling-3.0-flash, developed with financial institutions and domain experts. With 124B total and 5.1B active parameters, it supports information retrieval, research, valuation modeling and report preparation across long reports, research materials and complex workbooks. The model showed competitive results across FinFIRST, FinSearchComp Verified, FinCRAFT, FinanceAgent v1.1/v2, APEX-Agents, SpreadsheetBench v1/v2 and τ³-Banking. We will open-source the model weights next week.

Context
Ant Ling introduced Ling-3.0-flash-Fin on August 27, a finance-tuned version of its Ling-3.0-flash model that it says was developed with financial institutions and domain experts. The pitch is that financial work fails on small slips, an unreliable source, a figure defined two different ways or a wrong calculation, so a useful model has to retrieve the right documents, compute correctly and leave output a person can audit. The model keeps the base model's design: 124 billion parameters in total, of which only about 5.1 billion are used for any one , so each token takes far less computation than the full parameter count implies.
The model card on Hugging Face, from inclusionAI, describes the target tasks more concretely: searching that favors authoritative sources, reconciling conflicting figures across annual reports, earnings releases and regulatory filings, and building valuation spreadsheets with formulas that depend on other sheets. Ant calls its results competitive on finance and benchmarks including FinFIRST, FinSearchComp Verified, SpreadsheetBench and τ³-Banking, but neither the post nor the card text states the scores, and the card itself says the model still needs more validation on long, complex workflows and that its valuations and investment conclusions need professional review. Ant said the weights would follow the next week, and it open-sourced them on September 3 together with FinFIRST, its benchmark for financial search agents.
- mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
Checking sign-in…
Loading comments…






