Open-weight & owned stack · Quantization efficiency
Bonsai Publishes Ternary 27B Model Results
Bonsai published results for its Ternary Bonsai 2 27B model, reporting lower benchmark scores than the cited Qwen comparisons.
Read the original at reddit.comOpens the publisher's site in a new tabAlso covering this
PrismML’s Bonsai 2 27Bprismml.com, Sep 18PrismML releases Bonsai 2 27B, which compresses Alibaba's Qwen3.8 27B to 5.9 GB, small enough for smartphones, while retaining 98.2% of Qwen's benchmark scores (Julie Bort/TechCrunch)cnbc.com, Sep 18PrismML Squeezes a 27-Billion-Parameter AI Model Into 5.9 Gigabytes - Startup Fortunestartupfortune.com, Sep 17The 'Ternary Bonsai 2 27B' is an AI model that reduces the size of the Qwen3.8 27B to 5.9GB while maintaining 98.2% of its performance. - GIGAZINEprismml.com, Sep 18Ternary Bonsai 2 (27B) just released on Hugging Face. At <6GB in size, it can even run locally in-browser on WebGPU.reddit.com, Sep 17PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance - MarkTechPostmarktechpost.com, Sep 18Bonsai 2 27B Shrinks Local AI to 5.9GBlapaasvoice.com, Sep 18PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance - MarkTechPostmarktechpost.com, Sep 18