Qwen3.8 27B IQ3_XXS vs Bonsai Ternary PQ2 quantization test

A user benchmarked Qwen3.8 27B IQ3_XXS (10.18GiB) against Bonsai Ternary PQ2 (6.42GiB) on the same 16GB VRAM setup. Both completed 4/4 tasks, but Qwen ran 3.02x faster and used 3.10x fewer output tokens; Bonsai was slower with slightly worse results.

• Prism-ML Bonsai 2 Joins Qwen3.8 Quantization Comparison2026-09-19
• Builder assembles 768GB VRAM rig for under one RTX 60002026-09-18
• Qwen3.8 27b game dev post2026-09-16
• Voodoo Dynamic Quant goes MIT2026-09-15
• Hope for KVCache + Engram in Upcoming Models2026-09-14