PrismML Releases Ternary Bonsai 2 27B: 9x Smaller, 98.2% Performance Retained

Key Info

PrismML announced Ternary Bonsai 2 27B, a ternary quantized model based on Qwen3 27B that is 9x smaller than its full-precision counterpart while retaining 98.2% of aggregate benchmark performance, with the same 5.9 GB footprint as the original Bonsai 27B but significantly improved quality.

Highlights

• Two months after the first Bonsai 27B release, Ternary Bonsai 2 27B delivers materially narrower gaps to full precision, especially in agentic coding, multimodal reasoning, and long-horizon tool use.
• The model keeps the compact 5.9 GB footprint, making it practical for local deployment while achieving near-full-precision results.
• Based on Qwen3 27B, this release emphasizes quality gains rather than footprint reductions, positioning it as a stronger lightweight alternative for complex AI tasks.