PrismML Launches Bonsai 2 27B: 9x Smaller Than Qwen3.8, Keeps 98.2% Performance
ivan_bezdomny · x · 2026-09-19
PrismML open-sourced Ternary Bonsai 2 27B, a ternary-weight compression of Qwen3.8 27B that shrinks memory to 5.9GB while retaining 98.2% of aggregate benchmark performance, under Apache 2.0.
- Runs fully locally in-browser via WebGPU or on consumer GPUs like a 12GB RTX 3060 (220K context, 35 tokens/s decode measured)
- Targets agentic workloads: multimodal reasoning, agentic coding, and long-horizon tool use, with fewer silent failures and derailments on multi-step tasks vs the first 27B release
- Team demos local computer-use workflows on an RTX 5090
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
- 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
- Gemma.c: full Gemma 1/2/3 multimodal inference in a single C file, ~72 tok/s on CPU at INT8 — This-Purchase-3325 · 2026-11-16