Release of Ternary Bonsai 2 Model Suite
Prism ML has announced the release of Ternary Bonsai 2, a new suite of large language models (LLMs) available via the Hugging Face platform. The release includes three distinct model sizes: 2 billion (2B), 4 billion (4B), and 9 billion (9B) parameters. The collection is currently hosted in a dedicated repository on Hugging Face.
Performance and Technical Specifications
According to reports circulating on Reddit, the 2B parameter version of the model is designed for high-speed local execution. Unverified technical claims suggest that the 2B model can run entirely within a web browser using custom WebGPU kernels. On hardware configured with an M4 Max chip, users have reported performance speeds of up to 500 tokens per second.
A WebGPU demonstration of the Ternary Bonsai 2 Hermes model is currently available via Hugging Face Spaces to showcase these capabilities. While these performance metrics have been highlighted in community discussions, they lack independent, third-party verification at this time.
Community Reception and Discussion
The release has prompted varied reactions within developer communities. On Reddit, discussions have focused on the practical utility of the smaller 2B model. While some users questioned the specific use cases for a model of that scale, others engaged in speculative humor regarding its application in enterprise server optimization.
The release has not been met with universal acclaim. Some community members expressed skepticism regarding the utility of the models, with one user characterizing the speed-to-quality ratio as “garbage at the speed of sound.” Additionally, some commentary noted linguistic coincidences regarding the “Bonsai” name in different international contexts, which may influence the perception of the brand in certain regions.
As of this report, the models remain available for download and testing through the official Prism ML collection on Hugging Face.