Bonsai 27B mlx 1bit
1-bit quantized version of the 27B-parameter Qwen3.6-27B language model, using binary g128 weight representation (1.125 bits per weight) for a ~3.9GB deployed footprint, about 14.2x smaller than FP16. Supports a 262K token context, retains around 89.5% of FP16 accuracy across reasoning, math, coding, and tool-use benchmarks, and runs on-device on phones, laptops, and GPUs via Apple MLX and CUDA.
- Type
- Multimodal
- Weights
- Open
- Downloads
- 1M
- Likes
- 240
- Pipeline
- text generation
- License
- apache-2.0
Model hub
prism-ml/Bonsai-27B-mlx-1bit
- Repo
- prism-ml/Bonsai-27B-mlx-1bit
- Library
- mlx
- Access
- Public
Technical tags and languages
More from PrismML
13
-
16 DL
1 Bit Bonsai 1.7B
Text
1-bit Bonsai 1.7B is PrismML’s smallest Bonsai model, built for extremely fast on-device inference. With a me…
-
1 Bit Bonsai 4B
Text
1-bit Bonsai 4B is PrismML’s mid-size ultra-efficient model designed for fast local inference. It uses just 0…
-
1 Bit Bonsai 8B
Text
1-bit Bonsai 8B is PrismML’s flagship ultra-efficient language model, built with 1-bit weights for edge and r…
-
1-bit Bonsai Image 4B
Image
1-bit Bonsai Image 4B is PrismML’s open-weight compressed image-generation model built for local diffusion in…
-
1.1M DL
Bonsai 27B
Text
Bonsai 27B is a 27.8B-parameter multimodal language model built on Qwen3.6 27B, quantized end-to-end to 1-bit…
-
504 DL
Bonsai 27B AWQ 4bit
Multimodal
Bonsai 27B AWQ 4bit is a 4-bit AWQ quantized release of Prism ML's Bonsai 27B, a 27 billion parameter multimo…
-
1.2k DL
Ternary Bonsai 27B
Multimodal
A 27B parameter language model compressed to ternary (1.71 bit) weights, retaining 95% of full precision inte…
-
4k DL
Ternary Bonsai 27B AWQ 4bit
Multimodal
Ternary Bonsai 27B AWQ 4bit is a 4 bit AWQ quantized build of PrismML's Ternary Bonsai 27B, a multimodal mode…
About PrismML
1