PrismML's Bonsai 2 shrinks Qwen3.8 27B to 5.9GB, keeps 98.2% capability, runs on a 5090
dl_weekly · x · 2026-09-28
PrismML launched Bonsai 2 27B, an Apache-2.0 model that compresses Qwen3.8 27B from 56GB (16-bit) to 5.9GB using ternary weights (+1/0/-1), retaining 98.2% capability and running at 143 tokens/s on a consumer GeForce 5090. Benchmarks stay close to the original: agentic/tool-calling 77.6 vs 79.8, coding 81.6 vs 82.2, knowledge & reasoning 82.7 vs 81.3 — showing ternary compression can bring large models onto consumer PCs and high-end mobile devices.
More from Infra
- India's Sarvam launches Token Factory, hosting top open models on local infra — HarveenChadha · 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
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16