Open-weight & owned stack · Quantization efficiency
Underdog releases Saluki 27B quantized model
Underdog releases Saluki 27B, a roughly two-bit Qwen3.8-27B quantization fitting under 8GB.
Read the original at pc.watch.impress.co.jpOpens the publisher's site in a new tabAlso covering this
Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling - MarkTechPostmarktechpost.com, Oct 9Qwen3.8-27Bを8GBに凝縮！メモリ16GBのノートで動く「Saluki 27B」 | チバテレ＋プラスchiba-tv.com, Oct 9Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling - MarkTechPostmarktechpost.com, Oct 9Run Qwen3.8-27B Locally on Windows 11, Mac and Kali Linux with an 8 GB GPU: Underdog Saluki 27B GGUF (7.89 GB), Ollama and llama.cpp Install, Real RAM Requirements, GSQ-RCO Quants and AWS Costs | Logeshwaran.orglogeshwaran.org, Oct 10Underdog Saluki 27B: An 8GB Model That Beats Its 54GB Base on Tool Calls | MindStudiomindstudio.ai, Oct 13