Magnum v2 72B

Magnum v2 72B, developed by anthracite-org and based on Qwen2 72B, is positioned as a model aiming for high prose quality, particularly in roleplay scenarios, having been trained on 55 million tokens of curated RP data. In terms of overall performance, the model tends to have longer response times, ranking in the 19th percentile for speed across two benchmarks. Its pricing is moderate, placing it in the 20th percentile. Analyzing benchmark results, Magnum v2 72B demonstrates a notable weakness in General Knowledge, achieving only 40.5% accuracy (16th percentile), with a relatively high cost and duration for this category. Conversely, it shows a stronger performance in Email Classification, achieving 94.0% accuracy (27th percentile), indicating a reasonable ability to understand context and categorize emails effectively. This suggests a key strength in classification tasks, particularly those requiring nuanced understanding of text. Its primary weakness lies in broad general knowledge recall, where it struggles with both accuracy and efficiency. The model's reliability is not explicitly provided as a ranking, but the benchmark results suggest it consistently provides evaluable responses.