> ## Documentation Index > Fetch the complete documentation index at: https://docs.lighton.ai/llms.txt > Use this file to discover all available pages before exploring further. # Alfred Frontier 5 > LightOn introduces Alfred-ft5, the latest high‑performance model built for organizations that prioritize raw quality, advanced reasoning, and best‑in‑class tool usage. ## **A high-performance model, optimized for real-world enterprise workloads** Alfred-ft5 Alfred-ft5 is designed for teams that **want the very best performance** LightOn can deliver today. It is built on top of a modern **32-billion-parameter vision–language architecture**, **fine-tuned by LightOn** to excel in production RAG and agentic workflows. Beyond raw benchmark scores, Alfred-ft5 focuses on **robustness, reliability, and tool-centric behavior**: the model is trained and evaluated in realistic enterprise scenarios involving long contexts, multiple tools, and complex task decomposition. Where Alfred-sv5 prioritizes **European origin and sovereignty**, Alfred-ft5 targets organizations whose primary concerns are **quality and capability**. ## **Technical elements** In line with LightOn’s philosophy, the LLM is **one component** of a broader RAG and agentic stack. The real value lies in orchestration and retrieval. Still, for teams that need technical anchors, Alfred-ft5 offers: * a **context window of up to 32,768 tokens** enabling rich, multi-document prompts while keeping latency under control, * a **32B-parameter, vision–language architecture** capable of understanding both text and images for advanced multimodal workflows, * **strong instruction-following** behavior tuned for enterprise documentation, procedures, and compliance scenarios, * optimized **reasoning and planning** for multi-step tasks, including breaking down complex problems into actionable steps. LightOn's toolset ensures the context passed to the LLM is as high-quality as possible, with state-of-the-art parsing models and information retrieval stack, complementing these strengths and making sure Alfred-ft5 shines in RAG-centric and agentic use cases. ## **Best-in-class tool use and agentic behaviors** One of Alfred-ft5’s main differentiators is its **tool-calling performance**, particularly in **agentic setups powered by LightOn**: indeed, it offers improved **tool selection and argument construction**, reducing failed or ambiguous calls, as well as better **multi-tool coordination**, where the model must call several tools in sequence or in parallel. The model is **fine-tuned specifically to support and enhance LightOn’s tool ecosystem**. This includes RAG pipelines, function-calling interfaces, and higher-level agents, so that Alfred-ft5 behaves as a reliable, predictable decision-making core instead of a generic chat assistant. ## **Frontier vs Sovereign: two complementary product lines** LightOn structures its offering around two complementary model families: * **Alfred-ft5**: for organizations that prioritize **top-tier performance, reasoning, and tool use**, independent of model origin. * **Alfred-sv5**: for teams that must meet **European-origin** or similar sovereignty requirements, sometimes at the expense of using the single most capable global model. Both families benefit from the same **LightOn R\&D expertise and evaluation pipelines**. This separation lets organizations choose a model line that aligns with their **risk, compliance, and performance priorities**, while keeping the rest of their RAG and agentic stack unchanged.