LoRA-fine-tuned Qwen3-Coder-30B-A3B-Instruct specialised for the Crystal programming language. Quantized to Q8_0 GGUF for local inference via Ollama or llama.cpp.

This is v3 — a full-pipeline retrain on a substantially larger and cleaner corpus than v2 (crystal-qwen3.6-30b-gguf).

Quick start (Ollama)

bash

# Download the GGUF + Modelfile from this repo, then:
ollama create crystal-qwen-v3 -f Modelfile
ollama run crystal-qwen-v3 "Write a Crystal class Item with name and quantity that serializes to JSON."

What v3 changes vs v2

	v2 (`3.6`)	v3
CPT tokens	~3 M	~50 M (top-500 Crystal GitHub repos + stdlib + book + RFCs)
SFT pairs	~600	~3 000 (mined + Claude-Haiku-4.5 augmented, compile-gated)
DPO pairs	37	74 (chosen = idiomatic Crystal; rejected = Ruby-style or wrong)
LoRA rank / α	32 / 64	64 / 128
Quantization	Q4_K_M	Q8_0
Targets	attn only	attn + MLP + MoE `experts.gate_up_proj/down_proj`

Eval

Held-out eval (30 natural-language Crystal tasks; idiom score + crystal build --no-codegen compile gate). Higher is better.

Model	Idiom	Compile pass	Total
crystal-qwen-v3 (this model)	+76	26/28 (93 %)	+206
jaimef21/crystal-qwen3.6-30b (v2)	+67	21/28 (75 %)	+172
Qwen3-Coder-30B-A3B-Instruct (base)	+71	21/28 (75 %)	+176

v3 beats both prior baselines on every axis. v2 actually slightly trailed the base model on idiom (+67 vs +71) — v3 is the first checkpoint of this lineage that is unambiguously better than vanilla Qwen3-Coder at Crystal.

Caveat: v3 ships at Q8_0 while v2 and base measurements above are Q4_K_M, so some compile-gate fidelity gap is attributable to less quantization noise.

Files

crystal-qwen-v3-30b.gguf — 32 GB, Q8_0 (8.51 BPW)
Modelfile — Ollama Modelfile (chatml template + Crystal system prompt)
README.md — this file

Reproducing

Full reproduction pipeline (data scraping → training on RunPod H200 → GGUF quantization → eval) is documented at: github.com/jaimef/crystal-lora — see REPRODUCE.md.

License

Inherits Qwen's Apache 2.0 license from the base model.

jaimef21/crystal-qwen-v3-30b-gguf

crystal-qwen-v3-30b-gguf