PrismML launches ternary Bonsai 2 27B: 5.9GB, 9x smaller, retains 98.2% of full-precision performance
airesearch12 · x · 2026-09-18
PrismML launched Bonsai 2 27B, a new flagship built on Qwen3.8 27B using ternary weights. It fits in just 5.9GB — over 9x smaller than the full-precision model — while reportedly retaining 98.2% of aggregate benchmark performance (vendor-reported, not independently verified).
- The previous Ternary Bonsai 27B retained 95%, showing low-bit models are closing the quality gap quickly generation over generation
- Positioned for local deployment: reasoning, coding, vision, and long-horizon agentic workflows
- CEO Babak Hassibi: powerful models don't need to be confined to cloud infrastructure
The poster also notes his fully autonomous explainer-video bot @xplainervideo generated a video covering the launch, with quality steadily improving.
More from Models
- Numinous unveils Numinous-1, an 8B forecasting model fine-tuned on Qwen3-8B — const_reborn · 2026-09-18
- Grok Bot and Muse are fun but not smart enough for real-world work — jdjohnson · 2026-09-18
- Tencent-Backed AI Startup Valued at $1.42B to Release First Open-Weight LLM — kimmonismus · 2026-09-18
- Nathan Lambert: OpenAI hack via Claude shows closed models are the real AI risk tip — natolambert · 2026-09-18
- tokenbender: no benchmark can capture frontier models' inhuman blind spots in SWE/MLE — tokenbender · 2026-09-18
- Open-source models already at SOTA — Anthropic/OpenAI edge is just 5GW compute, dev argues — ccerrato147 · 2026-09-18