The new dale-0.8b-GGUF language model, released on Hugging Face, specializes in converting natural language queries into Linux Bash commands. It outputs results in JSON format with brief explanations. The model is distributed under the Apache 2.0 license, allowing for free use and modification.
In a benchmark of 471 tasks, dale-0.8B achieved a 79.4% success rate. It outperformed Qwen3.5-9B and Gemma 4 E4B, which both scored 62.8%, as well as Qwen3.5-4B at 54.6% and Qwen3.5-0.8B at 14.9%. These results were reported by the project author.
The model includes safety features to handle ambiguous or potentially harmful requests by returning refuse or unsure statuses. It is available in the GGUF format with Q4_K_M quantization, taking up approximately 529 MB of storage space.
Designed for local execution, the model runs on a CPU via llama.cpp, eliminating the need for a GPU. According to the developer, the median latency is 1.72 seconds on a 2020 laptop. It powers the dale command-line tool, which manages command suggestions and execution confirmation.
