Developers wanting a capable model on a phone or laptop get it free under Apache 2.0, though the free API is a limited-time preview.
Kenning's take
What it does
Bonsai 27B is a compressed version of Qwen3.6 27B from PrismML, using 1-bit and ternary weights. The 1-bit version is a 3.9 GB file that fits in an iPhone 17 Pro's memory, and the ternary version is 5.9 GB for laptops. Weights are free on Hugging Face under Apache 2.0, with a free limited-time preview API.
- Based on Qwen3.6 27B
- 1-bit version at 3.9 GB
- Ternary version at 5.9 GB
- Runs on Mac, iPhone and iPad via MLX and on NVIDIA GPUs
- Apache 2.0 licence
Who it suits. Developers who want to run a capable model on phones, laptops or small GPUs.
Key facts
- Category
- AI models
- Made by
- PrismML
- Released
- 14 July 2026
- How to get it
- Free weights on Hugging Face; free limited-time developer preview API through Together AI; usable in the Locally AI iOS app
- Works on
- Self-hosted, API, iOS (via Locally AI)
- Free plan
- Free open weights
- API price
- Free limited-time developer preview API; no paid pricing published
FAQ
When was Bonsai 27B released?
PrismML announced it on 14 July 2026.
Can Bonsai 27B run on an iPhone?
PrismML says the 1-bit version fits within the memory of an iPhone 17 Pro.
Is Bonsai 27B free?
Yes. The weights are under the Apache 2.0 licence, and there is a free limited-time preview API.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.