Underdog-Saluki-27B-1.0

What is Underdog-Saluki-27B-1.0

A 2-bit GGUF quantised model based on Qwen3.8-27B, with a size under 8GB. It focuses on preserving tool-calling capability and can be run directly with llama.cpp.

Key features

• A 2-bit GGUF quantised model based on Qwen3.8-27B, size 7.89GB
• Focuses on preserving tool-calling capability and supports parallel tool calls
• Supports text generation and can achieve image understanding via an optional vision add-on
• Thinking mode is enabled by default and can be toggled per request, consistent with Qwen3.8
• Compatible with stock llama.cpp for direct execution

Who it is for

Users who need to run tool-calling and multimodal tasks on large models but have limited hardware resources. Suitable for developers and researchers deploying llama.cpp locally.

How to get started

1. Run llama-server -m Underdog-Saluki-27B-1.0-IQ2-mix.gguf --jinja -ngl 99 -fa on -c 32768 to start the service
2. For vision capability, additionally download mmproj-Underdog-Saluki-27B-1.0-F16.gguf and load it with the --mmproj parameter
3. Send requests via the OpenAI-compatible chat API (port 8080), supporting tool calls and thinking control

Compiled from public descriptions on the tool's official website / README / launch page. Pricing and user numbers have not been verified; please check the official website. Some sections were drafted with AI assistance. Written in Chinese and English; other languages are machine-translated.