Large language models (LLMs) continue to evolve, demonstrating increasing capabilities in various fields. A recent test highlighted the performance of Little Qwen 3.5 27B and Qwen 35B-A3B models in logical reasoning.

Benchmark Details

The results were obtained using the lineage-bench benchmark and are publicly available. A key feature of these models is their ability to handle a large number of premises, opening new perspectives for applications that require complex reasoning.

Considerations

The ability of relatively small models to handle complex reasoning is a significant aspect. For those evaluating on-premise deployments, there are trade-offs between model size, hardware requirements, and performance. AI-RADAR offers analytical frameworks on /llm-onpremise to evaluate these aspects.