01 Public Benchmarks
Ranking #2 / #3 Publicly Available Models with 35B Parameters or Fewer
Taiwan-Aware, Validated by Public Benchmarks
AIEC validates FFM's understanding of Taiwan's language, culture, and local contexts.
AI Model Services | Model Foundry
AI Models Built for Taiwan
Accelerating AI Adoption
Taiwan AI Cloud's Formosa Foundation Model (FFM) is designed around Taiwan's linguistic and cultural context and enterprise application needs. From model evaluation and domain-specific optimization to knowledge integration and deployment, we help enterprises find the AI that best fits their needs.
Taiwan Context
Understand Taiwan's language, culture, and enterprise needs to deliver models tailored to local requirements.
Flexible Choices
From general-purpose and task-specific models to multimodal models, embeddings, and edge AI, choose the capabilities that best fit your needs.
more
General-Purpose Models
Language understanding, reasoning, and text generation
Agentic / Task-Specific Models
Reasoning and executing multiple tasks in parallel
Multimodal Models
Understanding images, text, and other content types
Embedding / Reranking Models
Improving the accuracy of internal knowledge retrieval
Lightweight / Edge Models
Low-latency inference and deployment in specialized environments
Compliance and Control
Select appropriate governance and deployment approaches based on data sensitivity and compliance requirements.
Why Model Foundry?
Choose the Right Model to Turn AI into Real-World Solutions.
Different Tasks, Different Models.
Different tasks - including Q&A, AI agents, document understanding, knowledge retrieval, and edge AI - require different model capabilities.
Balancing Performance and Cost Is Essential for Every Enterprise.
Although models with hundreds of billions of parameters offer powerful capabilities, training them in-house and deploying them on-premises require substantial computing power, memory, and server resources, driving up both initial infrastructure investment and long-term operating costs.
Data and Deployment Requirements Matter.
Data sensitivity, regulatory compliance, access permissions, and deployment environments directly influence model selection and integration approaches.
Skip to main content Choose Based on Task, Specifications, and Deployment Requirements
Model Performance Validation
More Than Taiwan-Aware: In-House Training Enhances Agent Capabilities and Deployment Flexibility.
02 In-House Training
Agent Task Capabilities Improved by Over 10%
From Understanding Language to Taking Action.
In-house model training further enhances reasoning, tool use, and multi-step task execution.
03 Sovereign AI Deployment
30B / 31B FFM Core Models
Better Suited for Domestic and On-Premises Deployment
Lower deployment barriers while balancing model capabilities, data security, inference efficiency, and long-term costs.
Source: AIEC official language model evaluation results (September 2026); agentic AI benchmarks based on PinchBench / OpenClaw Agent Benchmark. Data compiled as of September 30, 2026. Rankings may change as subsequent benchmark results are published.
Application Scenarios
Understands Taiwan, Understands Your Industry.
Enterprise Knowledge
Enterprise Knowledge Assistant
Integrates FAQs, product documentation, internal policies, and knowledge bases to support real-time queries, summarization, and internal enterprise Q&A.
Autonomous Execution / Task-Specific Models
Process Agent
It connects APIs, tools, and existing systems to perform data retrieval, task breakdown, cross-system processing, and workflow.
Multimodal Models
Document and Image Understanding
Understand forms, charts, scanned documents, and mixed text-and-image content to reduce the time spent on manual reading, organization, and information extraction.
Industry-Specific Applications
Financial Services
Financial Services | AI for Credit Assessment and Regulatory Compliance
Integrate supervision standards, internal operations, and credit information to help frontline personnel quickly find evidence, organize cases, and produce reports.
- Regulatory Requirements, Official Interpretations, and Internal Policy Queries
- Credit Application Data Consolidation and Risk Information Analysis
- Financial Data Summarization and Credit Report Drafting
Can Be Integrated with Enterprise Data
Regulatory InterpretationsCredit Assessment SOPsProduct SpecificationsRisk Documents
Manufacturing
Manufacturing | AI for Equipment Maintenance and Technical Knowledge
Integrates equipment and process knowledge to help engineering teams quickly identify root causes, find maintenance solutions, and access historical troubleshooting records.
- Fault Code Lookup and Troubleshooting Recommendations
- Search Equipment Manuals, Maintenance Records, and Process SOPs
- Maintenance Report and Technical Information Summarization
Can Be Integrated with Enterprise Data
Equipment ManualsMaintenance RecordsProcess SOPsFault Codes
Healthcare
Healthcare | AI for Clinical Knowledge and Medical Documentation
Provides a centralized access point for querying clinical documents and institutional knowledge, helping healthcare and administrative staff find the information they need more quickly.
- Clinical Guidelines and In-Hospital SOPs Lookup
- Summarization of Drug Information, Professional Documents, and Research Materials
- Cross-Document Knowledge Search and Information Synthesis
Can Be Integrated with Enterprise Data
Clinical GuidelinesIn-Hospital SOPsPharmaceutical DocumentsProfessional Reference Materials
Government/Public Sector
Government | AI for Regulatory and Administrative Knowledge
Integrates regulations, official interpretations, and administrative procedures to help public sector staff quickly find relevant legal references, organize documents, and address day-to-day operational inquiries.
- Cross-Document Search of Regulations and Historical Official Interpretations
- Search for Administrative Procedures, Forms, and Internal Policies
- Government Document Consolidation and Summarization
Can be combined with official documents
RegulationsOfficial InterpretationsAdministrative SOPsOfficial Correspondence and Documents
Compliance and Governance
Build AI for Governance from Day One, Not After Regulations Take Effect.
For highly regulated industries, addressing data retention, access controls, auditability, and deployment requirements only after implementation can lead to architectural redesign, data migration, and redeployment. Compliance should be built into the AI architecture from the design stage.
Data Can Remain Within Taiwan
Plan for in-country cloud or on-premises deployment based on project requirements and compliance needs, reducing governance and regulatory risks associated with cross-border data transfers.
Models Can Be Deployed in Private Environments
Run models in environments controlled by the enterprise, based on data sensitivity and industry requirements, minimizing unnecessary exposure of core business data to external parties.
Access Controls Can Be Managed
Control access to models, data, and knowledge bases based on user roles and use cases, giving enterprises greater control over how their data is accessed and used.
Auditable Usage
Maintain records required for model usage and governance to support the management needs of highly regulated sectors, including finance, healthcare, and government.
Model Selection
Complete Formosa Foundation Model (FFM) Series
01
Large Language Models (LLMs)
Enterprise Q&A, Long-Context Processing, Reasoning, and Tool Calling
Llama 3.3-FFM70B｜32KEnhance Tool Call
Llama 3.1-FFM405B / 70B / 8B｜32KEnhanced Traditional Chinese Support and Function Calling
Llama3-FFM70B / 8B｜8KEnhanced Traditional Chinese Support and Tool Calling
FFM-Mistral-7B7B｜32KLightweight Large Language Models (LLMs)
02
Multimodal Models
Image, OCR, and Document Understanding
FFM-Gemma4-31B31BAIEC-Evaluated Models
FFM-Muse-Glimmer-30B30BTMMLU+-Evaluated Models
Llama 3.2-FFM11B｜32KTraditional Chinese OCR and Image Understanding
03
Embedding Models
Enterprise Knowledge Bases and Semantic Search
FFM-Embedding-v2.11B｜8K｜2,048 dimensionsTraditional Chinese and Multilingual Semantic Search
FFM-Embedding-v2Embedding SeriesEnterprise Knowledge Retrieval
Deployment Options
Sensitive Data Shouldn't Have to Leave Your Controlled Environment to Benefit from AI.
Sending sensitive data directly to external public AI services may increase cross-border data transfer, outsourcing, and data governance risks. Highly regulated use cases should prioritize private cloud, in-country cloud, hybrid, or on-premises architectures.
PRIVATE / DOMESTIC CLOUD
Private Cloud / In-Country Cloud Deployment
- Data and models run in enterprise-controlled or in-country environments.
- Retain the flexibility and scalability of cloud computing.
- Balance Governance Requirements with Rapid Deployment
HYBRID / ON-PREMISE
Hybrid / On-Premises Deployment
- Keep sensitive data and core inference within the enterprise environment.
- Integrate with Existing Data Centers, Systems, and Access Control Architectures.
- Ideal for Highly Regulated Sectors, Including Finance, Government, and Healthcare.
Full-Stack AI Foundry
Build Sovereign AI That Is Controllable, Governable, and Continuously Evolving - From Models to Applications.
Taiwan AI Cloud's Model Foundry integrates AI computing and orchestration, data and security governance, model training and management, and AFS Suite and Agent applications to deliver end-to-end AI capabilities spanning compute, data, models, and applications.
01Computing Power
AI Infrastructure and Orchestration
02Data
Data and Security Governance
03Model
Training, Management, and Deployment
04Applications
AFS Suite and Agent