Average Ratings 0 Ratings

Total
ease
features
design
support

No User Reviews. Be the first to provide a review:

Write a Review

Average Ratings 0 Ratings

Total
ease
features
design
support

No User Reviews. Be the first to provide a review:

Write a Review

Description

Bonsai 27B stands as the latest multimodal flagship in the Bonsai lineup, marking the debut of a 27B-class model designed to operate on mobile devices. Built on the foundation of Qwen3.6 27B, it introduces an elevated level of capability for local devices, featuring advanced multi-step reasoning, structured tool interactions, vision tasks, and agentic loops for computer use that maintain coherence throughout multiple steps. The Bonsai 27B is available in two distinct variants. The Ternary Bonsai 27B employs ternary weights combined with FP16 group-wise scaling, achieving an effective weight of 1.71 bits and occupying a 5.9 GB footprint, suitable for high-performance laptop applications. In contrast, the 1-bit Bonsai 27B utilizes binary weights with identical group-wise scaling, resulting in an effective weight of 1.125 bits and a more compact 3.9 GB footprint, making it compatible with the memory constraints of devices like the iPhone 17 Pro. Both models operate seamlessly across the entire language network, including embeddings, attention mechanisms, MLPs, and the language model head, without resorting to higher-precision alternatives. They also feature a compact 4-bit vision tower, enabling on-device workflows to effectively interpret screenshots, documents, and camera inputs, enhancing user interaction and productivity. This innovative approach underscores Bonsai 27B's commitment to pushing the boundaries of mobile AI capabilities.

Description

Qwen3.5 represents a major advancement in open-weight multimodal AI models, engineered to function as a native vision-language agent system. Its flagship model, Qwen3.5-397B-A17B, leverages a hybrid architecture that fuses Gated DeltaNet linear attention with a high-sparsity mixture-of-experts framework, allowing only 17 billion parameters to activate during inference for improved speed and cost efficiency. Despite its sparse activation, the full 397-billion-parameter model achieves competitive performance across reasoning, coding, multilingual benchmarks, and complex agent evaluations. The hosted Qwen3.5-Plus version supports a one-million-token context window and includes built-in tool use for search, code interpretation, and adaptive reasoning. The model significantly expands multilingual coverage to 201 languages and dialects while improving encoding efficiency with a larger vocabulary. Native multimodal training enables strong performance in image understanding, video processing, document analysis, and spatial reasoning tasks. Its infrastructure includes FP8 precision pipelines and heterogeneous parallelism to boost throughput and reduce memory consumption. Reinforcement learning at scale enhances multi-step planning and general agent behavior across text and multimodal environments. Overall, Qwen3.5 positions itself as a high-efficiency foundation for autonomous digital agents capable of reasoning, searching, coding, and interacting with complex environments.

API Access

Has API Yes 

API Access

Has API Yes 

Screenshots View All

Screenshots View All

Integrations

APIFree No 
Alibaba Cloud Model Studio No 
BaseRT No 
Claw Code No 
Ollama No 
OpenClaw No 
Qwen No 
Qwen3.5-Plus No 
Together AI No 
ZooClaw No 

Integrations

APIFree Yes 
Alibaba Cloud Model Studio Yes 
BaseRT Yes 
Claw Code Yes 
Ollama Yes 
OpenClaw Yes 
Qwen Yes 
Qwen3.5-Plus Yes 
Together AI Yes 
ZooClaw Yes 

Pricing Details

No price information available.
Free Trial No 
Free Version No 

Pricing Details

Free
Open source
Free Trial No 
Free Version Yes 

Deployment

Web-Based No 
On-Premises No 
iPhone App Yes 
iPad App Yes 
Android App No 
Windows No 
Mac Yes 
Linux No 
Chromebook No 

Deployment

Web-Based Yes 
On-Premises Yes 
iPhone App No 
iPad App No 
Android App No 
Windows Yes 
Mac Yes 
Linux Yes 
Chromebook No 

Customer Support

Business Hours No 
Live Rep (24/7) No 
Online Support Yes 

Customer Support

Business Hours No 
Live Rep (24/7) No 
Online Support No 

Types of Training

Training Docs Yes 
Webinars No 
Live Training (Online) Yes 
In Person No 

Types of Training

Training Docs Yes 
Webinars No 
Live Training (Online) No 
In Person No 

Vendor Details

Company Name

PrismML

Founded

2026

Country

United States

Website

prismml.com/news/bonsai-27b

Vendor Details

Company Name

Alibaba

Founded

1999

Country

China

Website

qwen.ai

Alternatives

Alternatives

Claude Mythos Reviews

Claude Mythos

Anthropic
Bonsai Image Reviews

Bonsai Image

PrismML
Qwen3.6-27B Reviews

Qwen3.6-27B

Alibaba
Claude Opus 4.5 Reviews

Claude Opus 4.5

Anthropic