Chopsticks HQLarkquillAI Local v1: private, offline LarkquillAI that runs on your own computer.
| Model | Based on | Download | RAM | |---|---|---|---| | LarkquillAI Local 4B | Qwen3.5-4B (Qwen team, Apache-2.0) | 2.74 GB, 2 parts | 8 GB computer | | LarkquillAI Local 8B | Granite-4.2-8B (IBM, Apache-2.0) | 5.35 GB, 3 parts | 16 GB computer |
Easiest install (needs Ollama):
macOS / Linux: curl -fsSL https://raw.githubusercontent.com/ilikemacos/larkquillai/models/install/install.sh | sh -s -- 4b (or 8b)
Windows PowerShell: $env:LARKQUILL_SIZE='4b'; irm https://raw.githubusercontent.com/ilikemacos/larkquillai/models/install/install.ps1 | iex
Manual: download every .partN of one model, join them (cat part1 part2 ... > file.gguf on macOS/Linux, copy /b part1 + part2 file.gguf on Windows), and check the result against SHA256SUMS.txt. Then open it in LM Studio, llama.cpp or Ollama. Full instructions, credits and licenses: see the README.
Model weights are unmodified upstream Q4_K_M GGUFs; only the GGUF name/description metadata was changed. Apache License 2.0.