New flagship model brings 27B-class reasoning, coding, vision, and agentic capability into a dramatically smaller, faster, and more energy efficient deployment footprint
PASADENA, Calif., Sept. 17, 2026 /PRNewswire/ — PrismML today announced Bonsai 2 27B, its new flagship model based on Qwen3.8 27B. Available as a ternary model, Bonsai 2 27B brings a new generation of reasoning, coding, vision, and agentic capabilities to the Bonsai series.
The release follows PrismML’s first Bonsai 27B model, which demonstrated that a 27B-class multimodal model could be compressed enough to run efficiently on a local device. Bonsai 2 27B advances that milestone further, significantly improving model quality while preserving the dramatically smaller deployment footprint that defines the Bonsai series.
At just 5.9 GB, Ternary Bonsai 2 27B reduces memory footprint by more than 9x compared with its full-precision counterpart. However, while the original Ternary Bonsai 27B retained 95% of the aggregate benchmark performance of its full precision counterpart, the new Ternary Bonsai 2 27B retained over 98%. This considerable closing of the gap shows how quickly low-bit models are improving from one generation to the next.
“Bonsai 27B proved that powerful models do not have to be confined to cloud infrastructure,” said Babak Hassibi, Founder and CEO of PrismML. “With Bonsai 2 27B, we are closing the quality gap while keeping the same deployment advantages: a dramatically smaller footprint, strong local performance, and the ability to support more demanding workflows such as agentic coding, multimodal understanding, and long-horizon task execution.”
“AI deployment has historically forced a choice between model quality and the resources required to run it,” said Ion Stoica, Advisor to PrismML and Professor at UC Berkeley. “What is compelling about Bonsai 2 27B is how little capability is lost despite such a dramatic reduction in footprint. If that gap continues to close, it can fundamentally expand where highly capable models can be deployed.”
A flagship model for practical local AI
PrismML introduced its first Ternary Bonsai 27B model in July, compressing a 27B-class multimodal model by 9x. Ternary Bonsai 2 27B advances that goal further, significantly improving the tradeoff between model quality, and memory footprint.
This matters because 27B-class models increasingly do far more than answer questions. They reason through complex problems, work across text and visual inputs, write and debug code, call tools, and execute multi-step workflows. For these workloads, deployment is not only about fitting the model into memory. It is also about whether the model can run fast enough, and reliably enough to be useful in real applications.
Ternary Bonsai 2 27B is built for that broader deployment challenge. It is intended for local assistants, multimodal agents, private knowledge workflows, computer-use applications, and long-running agentic tasks that have traditionally depended heavily on cloud inference.
Strong capability retention at a fraction of the footprint
Across a 20-benchmark suite spanning reasoning, math, coding, instruction following, vision, and agentic tool use, Ternary Bonsai 2 27B achieves an aggregate score of 83.9, retaining 98.2% of Qwen3.8 27B’s performance of 85.4. The model reaches up to 143 tokens/second on NVIDIA GeForce RTX 5090.
These gains are central to PrismML’s “intelligence density” thesis: the future of AI will not be defined only by larger models in larger datacenters, but by how much useful capability can be delivered per unit of memory, compute, energy and deployment footprint.
Technical Details:
The Ternary Bonsai 2 27B model is a 27.8-billion parameter multimodal language model. It has been trained using Google v5 TPUs. It has been designed for seamless integration with existing AI workflows and is optimized for low-latency inference on consumer-grade CPUs, and edge GPUs.
Pricing and Availability:
Developers, researchers, and other users can download the Ternary Bonsai 2 27B models under the Apache 2.0 license for free starting September 17, 2026.
- Download Ternary Bonsai 2 27B model here
About PrismML:
PrismML is a U.S.-based artificial intelligence company focused on making AI more efficient and accessible. PrismML emerged from a team of Caltech researchers and was founded with support from Khosla Ventures, Cerberus, and Google, with continuing support from Samsung. For more information, visit the Website, LinkedIn, or X.
All registered trademarks and product identifiers belong to their respective corporate entities. Any other trademarks or product names referenced here are also owned exclusively by their relevant companies.
Contact: Gary Bird, PrismML, [email protected]
SOURCE PrismML
