Bonsai 27B: A 27B-Class Model that runs on a phone

Bonsai 27B, based on Qwen3.6 and available in ternary (5.9 GB) and 1-bit (3.9 GB) variants, is the first model of its capability class to run on a phone, enabling multi-step reasoning, tool calls, vision tasks, and agentic loops with a 262K-token context. Both variants use low-bit weights end-to-end across all network components, with the vision tower in compact 4-bit form, and support speculative decoding. The model is released under the Apache 2.0 license.