vsan/tiny-pickle-v3-coder-GGUF

1---2base_model: Qwen/Qwen3-Coder-30B-A3B-Instruct3library_name: llama.cpp4pipeline_tag: text-generation5license: apache-2.06tags:7- gguf8- qwen3-coder9- coding10- software-engineering11- moe12- q8_013- q4_k_m14- tiny-pickle15---16 17# Tiny Pickle v3 Coder — GGUF18 19Quantized GGUF releases of Tiny Pickle v3 Coder.20 21Tiny Pickle v3 Coder was produced by fine-tuning22`Qwen/Qwen3-Coder-30B-A3B-Instruct` with the LoRA adapter published at23`vsan/tiny-pickle-v3-coder-LoRA`, then merging and converting the resulting model with24llama.cpp.25 26## Files27 28| File | Quantization | Approximate size |29|---|---|---:|30| `tiny-pickle-v3-coder-q8_0.gguf` | Q8_0 | 31G |31| `tiny-pickle-v3-coder-q4_k_m.gguf` | Q4_K_M | 18G |32 33Q8_0 retains greater numerical fidelity but requires more storage and34memory. Q4_K_M is smaller and more practical for local inference.35 36## Run with llama.cpp37 38```bash39llama-cli \40  -m tiny-pickle-v3-coder-q4_k_m.gguf \41  -ngl 99 \42  -c 8192 \43  -p "Write a robust Python LRU cache with unit tests."44```45 46## Intended use47 48- Code generation49- Debugging50- Code review51- Implementation planning52- Test generation53- Software-engineering assistance54 55## Limitations56 57Tiny Pickle v3 Coder is experimental and has not yet been proven superior58to its base model on independent benchmarks. Quantization may reduce model59quality. Generated code can be incorrect, insecure, incomplete, or60non-functional and must be reviewed and tested.61