Ant Ling today launched the finance-enhanced model Ling-3.0-flash-Fin. It retains the model architecture and long-context capabilities of Ling-3.0-flash, with the same configuration of 124B total parameters and 5.1B active parameters. The Ling-3.0-flash-Fin model weights will officially be open-sourced next week.
Building on this foundation, the model uses continued pretraining on financial corpora, domain-specific post-training, and tool-use optimization to further strengthen its ability to process complex financial content such as annual reports, financial workbooks, and multiple research materials, while also maintaining deployment efficiency.
According to the official announcement, Ling-3.0-flash-Fin was tested on financial agent benchmarks including FinFIRST, jointly developed by Ant Group and CICC, as well as FinSearchComp Verified, Finance Agent, APEX-Agents, SpreadsheetBench, and τ³-Banking. These tests covered scenarios such as financial information retrieval, investment research and analysis, long-horizon financial task execution, valuation modeling, and banking applications. The results show that Ling-3.0-flash-Fin demonstrates strong competitiveness on these tasks compared with several general-purpose models.
While its financial capabilities have been strengthened, the model's score on the AA Intelligence Index v4.1.1 also increased from 38 to 41, delivering coordinated improvements in both general and specialized capabilities.
Ant Group and CICC jointly developed FinFIRST (Financial Information Retrieval, Sourcing and Traceability). The benchmark was designed with in-depth participation from more than 50 financial professionals. It systematically evaluates answers across three dimensions: results, process, and evidence. This includes whether answers are accurate and complete, whether the research methodology and calculations are sound, and whether key data and conclusions can be traced back to authoritative primary sources. FinFIRST will be open-sourced soon, providing a reliable benchmark for financial search research and evaluation.
Evaluation results show that Ling-3.0-flash-Fin's overall score has surpassed those of some large flagship models. Compared with models of a similar size, it is highly competitive, particularly in prioritizing official, primary, and highly credible sources.
For financial professionals and developers, Ling-3.0-flash-Fin will be available on OpenRouter with a one-month limited-time offer of free API calls, allowing users to quickly experience and validate the model's capabilities in financial research, data analysis, application development, and other scenarios.
OpenRouter access:
https://openrouter.ai/inclusionai/ling-3.0-flash-fin:free #providers