PrismML, a company founded by Caltech researchers, has released Bonsai 27B, a 27-billion-parameter AI model designed to run directly on an iPhone. The model is based on Alibaba's Qwen3.6-27B and supports multistep reasoning, tool use, image understanding, and agent-based tasks. PrismML argues that modern AI apps increasingly need powerful models to run locally. An agent may make hundreds of model calls in sequence, each carrying context, producing structured output, and feeding into the next step. In the cloud, per-token costs pile up, every call adds network latency, and intermediate results, tool calls, and private data such as screen content or documents all leave the device. But running the model on-device cuts the marginal cost of those loops to zero and keeps user data local. PrismML sees this as the basis for always-on agents, offline assistants, and hybrid systems. Simple and privacy-sensitive tasks stay on-device, while only the hardest steps are sent to frontier models in the cloud. According to a CNBC report, PrismML is already in talks with Apple about the compression technology behind Bonsai. PrismML CEO Babak Hassibi confirmed that Apple and other companies are testing the models for speed, power draw, and performance. The talks are 'very early,' but 'things are progressing nicely.'
A model this size typically takes up about 54 GB of storage. Even with standard compression, it still needs around 18 GB. PrismML offers two much smaller versions: The quality-focused variant takes up about 5.9 GB and is meant for laptops, though the packages currently shipping may be larger depending on the runtime. The white paper lists about 7.2 GB for the llama.cpp version and 8.49 GB for the MLX version. The smaller variant comes in at about 3.9 GB, small enough to fit within the limited storage of an iPhone 17 Pro Max. According to PrismML, an iPhone with 12 GB of RAM actually makes only about 6 GB available to a single app, split between the model and the cache. Instead of storing each neural network weight as 16 bits, PrismML uses only one or just under two bits. In the most aggressive variant, each weight has only two states. In the slightly larger one, three. This approach is applied across the entire language model. As an example of common labeling issues, PrismML points to the Qwen3.6-27B-IQ2_XXS build compared in the white paper, which averages 2.8 bits per weight despite its '2-bit' label. The 1-bit Bonsai variant leads in efficiency with an intelligence density of 0.530 per GB, far ahead of ternary and FP16 models.
PrismML says compression has a limited impact on quality. In PrismML's own evaluation across 15 benchmarks, the larger variant keeps 95 percent of the original model's performance. The smaller one keeps 90 percent. Math and coding stayed 'virtually unaffected,' according to PrismML. The bigger drops showed up with the more aggressive compression, especially in image understanding, instruction following, and agent-based tool use. A conventionally compressed Qwen3.6-27B model at 9.4 GB scores only 72.7 points, while the smaller Bonsai variant at 3.9 GB scores 76.1. The compressed Bonsai models retain up to 95 percent of Qwen3.6-27B's original performance. The 1-bit variant drops off more sharply in vision and instruction following. Source: thedecoder
