Summary

Bonsai 27B is PrismML's low-bit multimodal language model family based on Qwen3.6 27B and designed to run locally on consumer hardware, including phones. The release includes ternary and 1-bit variants that preserve a large-model capability tier while reducing memory requirements enough for laptops, GPUs, and phone-class deployment.

Key Features

Runs a 27B-class multimodal model in low-bit form on local devices.
Offers a ternary 27B variant at about 5.9 GB for laptop-class quality.
Offers a 1-bit 27B variant at about 3.9 GB for phone-class memory budgets.
Uses low-bit representation across embeddings, attention, MLPs, and the LM head.
Includes a compact 4-bit vision tower for image, document, screenshot, and camera workflows.
Supports a 262K-token context window.
Supports speculative decoding for draft-and-verify acceleration.
Ships model weights under the Apache 2.0 license with local and API access paths.

About Bonsai 27B


The ternary variant uses 1.71 effective bits per weight and targets laptop-class quality at about 5.9 GB, while the 1-bit variant uses 1.125 effective bits per weight and fits around 3.9 GB. Both variants keep the low-bit representation across the language network, support a compact 4-bit vision tower, offer a 262K-token context, and support speculative decoding.


Bonsai 27B is useful for private local assistants, offline agentic workflows, hybrid cloud-local systems, on-device vision use cases, and cost-sensitive agent loops. PrismML reports strong retention of full-precision baseline performance across math, coding, tool calling, instruction following, knowledge, and vision benchmarks, with Apache 2.0 weights available on release.

Get more likes & reach the top of search results by adding this button on your site!

Embed button preview - Light theme
Embed button preview - Dark theme
TurboType Banner

Subscribe to the AI Search Newsletter

Get top updates in AI to your inbox every weekend. It's free!