null
Ant Group today announced the open-sourcing of Ling-3.0-flash-Fin at the 2026 Inclusion·Conference on the Bund. Grounded in actual business needs as an open model for real-world financial workflows, this release continues Ant Group's exploration of bringing AI into practical scenarios, focusing on delivering utility and high efficiency for complex tasks.
Solving Real-World Challenges Through Efficiency
Financial work depends on trustworthy sources, consistent definitions, accurate calculations and auditable outputs. AI in this sector must go beyond simple question-and-answer interactions to navigate dynamic information, complex accounting standards, and stringent compliance requirements. Ling-3.0-flash-Fin was co-developed with leading financial institutions and industry experts to embed professional expertise directly into task definition, data systems, and evaluation frameworks, guided by the principles of being professional, efficient, and open.
Built on a highly efficient Mixture-of-Experts (MoE) architecture, Ling-3.0-flash-Fin features 124 billion total parameters but activates only 5.1 billion per token. This design delivers the comprehensive knowledge capacity of a large model while maintaining the low inference costs and high deployment efficiency of a compact model.
This architectural efficiency translates into strong practical performance, with the model showing competitive results across various benchmarks including FinFIRST, FinSearchComp Verified, FinCRAFT, FinanceAgent v1.1/v2, APEX-Agents, SpreadsheetBench v1/v2, and τ³-Banking.
Core Capabilities for Professional Workflows
Ling-3.0-flash-Fin focuses on four core capabilities tailored for investment research and financial analysis:
·Information Retrieval: Prioritizes authoritative, official sources to ensure data consistency and end-to-end traceability.
·Research Reasoning: Synthesizes multi-source, heterogeneous data to construct clear, verifiable logical evidence chains.
·Valuation Modeling: Understands complex Excel financial linkages, supporting automated updates while keeping files fully editable.
·Report Generation: Integrates facts, calculations, and visual charts to produce highly readable, professional research outputs.
Now, Ling-3.0-flash-Fin is available on OpenRouter and Vercel and its open weights can be accessed through Hugging Face and ModelScope. Users can deploy it privately, connect it to search, Python, databases and spreadsheets, and adapt it to their own financial workflows.
Alongside the model, Ant Group is also open-sourcing FinFIRST, an expert-built benchmark for financial search agents. Developed with professional support from the investment banking team at China International Capital Corporation Limited (CICC), FinFIRST V1 includes 123 expert-authored tasks, 701 atomic criteria, and 12,300 rubric points. By evaluating the full research process rather than relying solely on final-answer matching, FinFIRST ensures strict data consistency and end-to-end traceability in complex financial scenarios.
A Broader Ecosystem for Diverse Needs
Recognizing that different industries and scenarios require specialized solutions, the Ling 3.0 series offer a rich portfolio of models beyond Ling-3.0-flash-Fin.
·Ling-3.0-flash: A hybrid-reasoning MoE model built for production-scale agents, featuring 124B total parameters with 5.1B active per token. It achieves performance comparable to flagship 1T models on most benchmarks while reducing compute requirements significantly.
·Ling-3.0-tiny: A native hybrid reasoning model with 7.9B total parameters and 1.3B active per token. Designed for resource-sensitive deployment, it runs entirely locally without cloud dependency, ideal for personal knowledge management and offline tasks.
·Ling-3.0-flash-VL: A vision-language model built on Ling-3.0-flash, adding image and video inputs with a 1M token context. It is designed for visual perception, STEM reasoning, document intelligence, multimodal agent tasks, and medical report interpretation.
·Ling-3.0-flash-Santé: An MoE model enhanced for healthcare and life sciences, achieving top-tier performance among flash-size models on key medical benchmarks like MedXpertQA-Text and DiagnosisArena-MCQ, delivering flagship-level performance for clinical reasoning and evidence-based retrieval.
