Bonsai 2 27B: 9x smaller ternary model keeps 98.2% of benchmarks, runs on a Mac
MaziyarPanahi · x · 2026-09-18
PrismML released Ternary Bonsai 2 27B, built on Qwen3.8 27B with {-1,0,+1} ternary weights and FP16 group-wise scaling (1.76 effective bits/weight). The model is just 5.9GB — over 9x smaller than its full-precision counterpart — while retaining 98.2% of aggregate benchmark performance. It supports 262K context, multimodal image+text input, Apache 2.0 license, with stronger reasoning, coding, vision and agentic capability. One user ran it locally on a Mac Studio to read medication lists in 5.7 seconds.
Related event: PrismML's Ternary Bonsai 2 27B Shrinks to 5.9GB, Retains 98.2% Performance(20 posts)→
More from Infra
- Training-free page-sparse decode in llama.cpp: 1.72× faster 128k decoding on an RTX 5090 — topppggg · 2026-11-16
- AMD MI455 beats Nvidia Rubin on raw specs, claims viral chip debate — zephyr_z9 · 2026-11-16
- llama.cpp offloads CI caches and runners to Hugging Face to survive 35 PRs/day — ggerganov · 2026-11-16
- India's Sarvam launches Token Factory, hosting top open models on local infra — HarveenChadha · 2026-11-16
- ZGX Fury stress-tested: 351 tok/s on Qwen3.8-Flash-Next, offline agents and local RAG demos — Sam Witteveen · 2026-11-16
- Gemini 3.5 Flash now runs on air-gapped Google Distributed Cloud, offline up to 12 months — shashib · 2026-11-16