Prime Intellect: INTELLECT-3

Prime Intellect's INTELLECT-3, a 106B-parameter Mixture-of-Experts model, demonstrates exceptional performance for its size. It consistently ranks among the fastest models and offers highly competitive pricing across all benchmarks. With a 97% success rate, its reliability is notably high, indicating minimal technical failures. The model exhibits significant strengths in specialized domains. It achieves a remarkable 96.0% accuracy in Mathematics, placing it in the 92nd percentile, and a strong 91.0% in Coding (68th percentile). General Knowledge and Ethics also show high accuracy at 98.0% for both. However, INTELLECT-3 struggles significantly with tasks requiring it to acknowledge uncertainty, scoring 0.0% on Hallucinations, and similarly on complex Instruction Following with 0.0% accuracy. Its Reasoning capabilities are moderate at 52.0% accuracy (29th percentile). While its cost and duration for most tasks are competitive, the duration for Mathematics and Reasoning benchmarks is notably high.