Agentica: Deepcoder 14B Preview

Agentica: Deepcoder 14B Preview, a 14B parameter code generation model, demonstrates a balanced performance profile with notable strengths in cost-efficiency and reliability. While its speed performance is moderate, ranking in the 21st percentile across benchmarks, it consistently offers highly competitive pricing, placing in the 89th percentile. The model exhibits strong reliability with an 87% success rate, indicating consistent and usable responses. In terms of specific benchmarks, Deepcoder 14B Preview excels in Instruction Following, achieving perfect accuracy in one instance, showcasing its precision. It also performs strongly in Coding, with 84.7% accuracy, and General Knowledge (84.0% accuracy), though its percentile rankings in these categories suggest room for improvement relative to top performers. Its primary weakness lies in Reasoning, where it achieved only 42.0% accuracy, placing it in the lower quartile. Despite some lower percentile rankings in accuracy for certain categories, its cost-effectiveness across most benchmarks is a significant advantage.