What is PrismML Ternary Bonsai 2 27B?

Large language models are programs trained on a vast amount of text so they can generate, summarize, translate and analyze writing. Their many stored parameters take a lot of memory, so one way to shrink them is quantization, which lowers the precision of those parameters while trying to keep most of the model's performance. Ternary Bonsai 2 27B is a language model from PrismML built on Qwen3.8 27B (an existing language model from Alibaba's Qwen team) that stores its weights as just three values (-1, 0 and +1) with a shared scaling factor for each small group of weights. It accepts text and image input, and the maker describes its Bonsai 2 family as bringing multimodal and agentic capabilities to local AI applications on personal computers, smartphones and wearables. It is released under the Apache 2.0 license.

Written from primary sources and reference pages, and checked against them by an independent AI fact-checker. Drafted with AI assistance.