PrismML: Ternary Bonsai 2 27B

PrismML's Ternary Bonsai 2 27B, a 27B-parameter reasoning model derived from Qwen3.8-27B, demonstrates exceptional speed and competitive pricing. It consistently ranks among the fastest models and offers highly competitive pricing across all benchmarks. The model also exhibits strong reliability, achieving a 94% success rate, indicating consistent delivery of usable responses. In terms of performance across categories, Bonsai 2 27B shows significant strengths in several areas. It achieves high accuracy in Email Classification (99.0%, 87th percentile), Reasoning (96.0%, 78th percentile), and Ethics (99.0%, 53rd percentile), making it well-suited for tasks requiring logical deduction, ethical judgment, and categorization. Its General Knowledge is also robust at 98.5% accuracy (51st percentile). However, a notable weakness is its performance in Hallucinations, with only 54.0% accuracy (9th percentile), suggesting it frequently fails to acknowledge uncertainty when presented with fictional information. Furthermore, its Instruction Following capability is currently at 0.0% accuracy, indicating a significant area for improvement. Despite these limitations, its overall speed and cost-effectiveness, combined with strong performance in core reasoning and classification tasks, position it as a compelling option for specific applications.