PrismML releases Ternary Bonsai 2 models for local AI on phones and wearables
They join the existing Bonsai 2 27B model, expanding options for developers targeting limited-resource devices like smartphones and wearables.
Key points
- PrismML’s **Ternary Bonsai 2** models (2B, 4B, 9B) use ternary weights to cut memory to **0.43GB–2GB**, enabling local AI on phones and wearables
The new models prioritize practical tasks such as instruction following, tool use, and concise responses. Benchmarking across 19 tests in six categories—agentic tool use, math, instruction following, coding, knowledge, and reasoning—shows improved performance over their full-precision counterparts. The models support NVIDIA GPUs (via llama.cpp) and Apple devices (via MLX), with weights available under Apache 2.0.
Ternary Bonsai 2 Family: 2B, 4B, & 9B based on Qwen 3.5
prismml.com · 15 October 2026.png)
Loading the full article…
This text was published by prismml.com and written by PrismML. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- OpenAI introduces Dots agent for asynchronous task delegation · 4 src
- Gmail Live rolls out to paid Workspace users · 1 src
- Meta's Muse AI permits explicit adult conversations, unlike other Meta apps · 4 src
- Anthropic adds OpenTelemetry trace export to Claude Platform for debugging and monitoring · 3 src
- Perplexity open-sources SPACE, its agent sandbox platform · 2 src
Comments
via GitHub Discussions