Liquid: LFM 3B

Liquid's LFM 3B, launched on January 25, 2025, is positioned as a high-performance model for its size, specifically targeting mobile and edge text-based applications. It demonstrates competitive response times, ranking in the 47th percentile across various benchmarks, and offers exceptionally competitive pricing, placing in the 96th percentile. In terms of benchmark performance, LFM 3B exhibits a notable strength in hallucination mitigation, achieving 86.0% accuracy, making it the most accurate model at its price point for this specific task. It also performs reasonably well in Email Classification with 93.0% accuracy. However, the model shows significant weaknesses in other areas. Its accuracy is very low for General Knowledge (0.8%), Ethics (6.0%), Instruction Following (24.0%), Reasoning (24.0%), and Coding (6.0%). While its pricing remains highly competitive across all benchmarks, its duration for some tasks, such as General Knowledge and Coding, is quite high. Overall, LFM 3B excels in cost-effectiveness and its ability to avoid hallucinations, but its general knowledge, ethical reasoning, and coding capabilities require substantial improvement.