Launch of Bonsai 27B Model Capable of Running on Smartphones
Bonsai 27B is a new AI model that can run on smartphones, reducing the model size to 3.9 GB. It is available in two open-source versions.
Launch of Bonsai 27B Model Capable of Running on Smartphones
PrismML has introduced Bonsai 27B — a 1-bit and ternary version of Qwen3.6-27B.
Thanks to the low-bit representation, the size of the original FP16 model has been reduced from 54 GB to 3.9 GB — approximately 93%, or 14 times smaller.
Available Versions:
1-bit Bonsai 27B — 3.9 GB
Optimized for running on smartphones and uses 1.125 effective bits per weight.
Ternary Bonsai 27B — 5.9 GB
Designed for laptops and uses 1.71 effective bits per weight.
The ternary version retains about 95% of the model's performance with full precision, while the 1-bit version retains about 90%. The evaluation covers reasoning, mathematics, programming, following instructions, using tools, and working with images.
Both versions are published as open-source under the Apache 2.0 license.
Thanks to custom WebGPU cores developed using Fable 5 and GPT-5.6 Sol, Bonsai 27B can run locally right in the browser — without sending data to the cloud.
Why it matters
AnalysisThis model opens new possibilities for using AI on mobile devices, making technology more accessible. Reducing model size without sacrificing performance is an important step in AI development.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
MIT researchers show GPT-6 Astra can improve other AI models by editing their weights
In an experiment, Astra raised Qwen3.5-0.8B's accuracy on MATH-500 from 7.9% to 48.1% by changing only 0.016% of the model's weights.
JSON — Structured Data for Programs, APIs, and AI Agents
A complete hands-on guide to JSON: syntax specifications, serialization via parse and stringify, REST API consumption, JSON Schema validation, and structured outputs for LLMs.
DOCX and XLSX Formats: When AI Needs to Deliver Production-Ready Documents
A comprehensive hands-on guide to generating and processing DOCX documents and XLSX workbooks with AI: comparing against Markdown and CSV, prompt engineering, formula design, and code automation.
Binance publishes investigation into $6.3 billion theft
Binance’s investigation links the theft to a modified Homebrew download, a fake interview and abuse of an internal AI tool.