Google: Gemma 4 31B

Google's Gemma 4 31B Instruct model demonstrates a strong overall performance profile, particularly excelling in reliability and cost-effectiveness. It consistently provides usable responses, achieving a perfect 100% success rate across all benchmarks, indicating exceptional stability. The model offers competitive pricing, ranking in the 73rd percentile for cost across various tasks. While its speed performance is competitive, ranking in the 41st percentile, it is not among the fastest models available. Key strengths include perfect accuracy in Hallucinations (Baseline) and Ethics (Baseline), where it also stands out as the most accurate model at its price point and speed. It shows impressive accuracy in Instruction Following (93rd percentile), Email Classification (90th percentile), and Mathematics (95th percentile), often being the most accurate among models of comparable speed. The model also performs very well in Coding (89th percentile). Its ability to handle a 256K token context window, support multimodal input, and offer configurable reasoning modes further enhances its versatility. While its General Knowledge and Reasoning scores are solid, they are not as high-ranking as its other top performances.