Written by Jakub Rusinowski · Last updated July 21, 2026
PREVIEW (July 2026). The first 27B-class model that runs on a phone. Bonsai 27B is PrismML's extreme low-bit rebuild of Qwen3.6-27B: the architecture is unchanged, but the weights are squeezed to roughly 1–1.7 bits each. The 1-bit build is just 3.9 GB — small enough for an iPhone — while the ternary build is 5.9 GB and aimed at laptops. Both are multimodal (text + images), carry a 262K-token context, and ship under Apache 2.0. PrismML reports the ternary build keeps more than 95% of the full-precision (FP16) baseline across a 15-benchmark suite and the 1-bit build more than 90%, running about 11 tok/s on an iPhone 17 Pro. Specs are from launch coverage — verify against the Hugging Face model cards.
| 1-bit Bonsai 27B | Min 4 GB VRAM · 1-bit (binary, ~1.125 bpw) · 262,144 ctx · |
| Ternary Bonsai 27B | Min 6 GB VRAM · Ternary (1.58-bit, ~1.71 bpw) · 262,144 ctx · |
The cheapest GPU that runs Bonsai 27B locally (min 4 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run
Minimum VRAM: 4 GB. For best results use Q4_K_M quantization.
Bonsai 27B needs about 4 GB VRAM at 1-bit (binary, ~1.125 bpw) quantization for its smallest variant. Variants: 1-bit Bonsai 27B (4 GB, 1-bit (binary, ~1.125 bpw)); Ternary Bonsai 27B (6 GB, Ternary (1.58-bit, ~1.71 bpw)). On Apple Silicon, unified memory counts toward this requirement.
Yes — Bonsai 27B runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at 1-bit (binary, ~1.125 bpw).
Q4_K_M is the best balance of quality and VRAM for Bonsai 27B in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run . This downloads Bonsai 27B and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.