Skip to main content

​
A European model, optimized for enterprise workloads

Alfred-sv5 Alfred-sv5 is built on a modern 24-billion-parameter architecture, fine-tuned by LightOn to better align with the retrieval and orchestration pipeline. The result is improved stability, more reliable outputs on enterprise data, and seamless integration within search, retrieval, and enterprise agent workflows. Compared to Alfred 4.2, Alfred-sv5 5 improves upon an issue that could occasionally cause instabilities or looping behaviours. It now offers significantly enhanced stability, particularly for long, complex or highly structured queries.

​
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 retrieval and orchestration. Still, for teams who need technical touchpoints, Alfred-sv5 offers:
  • a context window of up to 32,768 tokens enabling rich, multi-document prompts while keeping latency under control,
  • improved instruction-following capabilities and reduced repetitive or looping outputs,
  • strong support for function calling and structured outputs, enabling tighter integration with tools, agents, and enterprise systems.
These capabilities complement rather than replace the upstream work: indexing, embeddings, retrieval, filtering, and context composition.

​
Frontier vs Sovereign: two complementary product lines

Alfred-sv5 addresses organizations that specifically require a European-origin LLM, even when other open-source models already provide complete confidentiality. 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.

​
A component of a RAG system, not the destination

In a fully operational RAG system, the LLM is just one component among many: indexing, embeddings, retrieval, guardrails, agent workflows, and post-processing all play decisive roles.
Alfred-sv5 positions itself as the sovereign core model within this broader pipeline without pretending to be the sole driver of system performance. This combination—a well-designed RAG pipeline, a model fine-tuned for enterprise use, and a choice between sovereign or high-performance options—allows organizations to balance political requirements, regulatory constraints, and real-world impact.