Bonsai Models by PrismML
Last refreshed 2026-10-08. Next refresh: weekly.
Details
Capabilities
About
Bonsai is PrismML's family of 1-bit and ternary open-weight language models for local / edge deployment (docs: 'Bonsai is a family of 1-bit and ternary language models from PrismML'). Line includes Ternary Bonsai 2 27B (2026-09-17; based on Qwen3.8 27B), prior Bonsai 27B generation (2026-07-14; based on Qwen3.6 27B — Ternary + 1-bit variants), smaller Bonsai 8B/4B/1.7B, and Bonsai Image. Ternary Bonsai 2 27B is the current flagship, with prior Ternary Bonsai 27B as a sibling under the same family.
Decision facts
- Best fit
- agentson devicevision and multimodal work
- Capability starting point
- Ternary Bonsai 2 27B with 256K context and reasoning, JSON / Tool use, and multimodal inputs
- Lowest tracked input
- Not tracked
- Closest related family
- Agents-A1
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs on device, 256K context, and 27.4B parameters.
Use when the workload needs on device, 256K context, and 27B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Ternary Bonsai 2 27B | Use when the workload needs on device, 256K context, and 27.4B parameters. | 2026-09 | on device256K context27.4B parameters | Current |
| Ternary Bonsai 27B | Use when the workload needs on device, 256K context, and 27B parameters. | 2026-07 | on device256K context27B parameters | Current |
Release Timeline
2 release groupsSpecifications(2 models)
| Model | Released | Context | Parameters | Vision | Multimodal | Reasoning | JSON / Tool use |
|---|---|---|---|---|---|---|---|
| Ternary Bonsai 2 27B | 2026-09 | 256K | 27.4B | Yes | Yes | Yes | Yes |
| Ternary Bonsai 27B | 2026-07 | 256K | 27B | Yes | Yes | Yes | Yes |