---

license: apache-2.0

base_model: Qwen/Qwen3-4B-Instruct-2507

language:

• de
• ar

pipeline_tag: text-generation

tags:

• gguf
• qwen3
• german
• language-learning
• on-device

---

Laura — German tutor in Arabic (Qwen3 4B Instruct 2507, GGUF)

Laura is the tutor in an offline German course app for adult beginners

(A1–A2). This model talks to the learner in Arabic and teaches German with

German examples. It runs on the phone, CPU only.

This is a modified version of Qwen/Qwen3-4B-Instruct-2507.

Changes made: a LoRA adapter (rank 16, all linear layers, 2 epochs) was trained

on ~2,500–3,300 short tutoring conversations in Arabic, merged into the weights,

and the result converted to GGUF and quantised to Q4_K_M with llama.cpp.

Distributed under the base model's Apache-2.0 licence.

How the app uses it

The app does not trust the model to judge German on its own. It checks the

learner's sentence with its own dictionary first and adds the verdict to the

message — (Check: ✗ → „corrected sentence“ · reason) or (Check: ✓ …) — plus

dictionary meanings and the chapter's rule card. The model was trained with

exactly those notes and explains them; without them it is a much weaker teacher.

Evaluation

Pro tutor in Arabic, 4B (T-119), round 8 (1 Oct): Qwen3-4B-Instruct-2507 with a LoRA (r16, 3 epochs) on the Arabic tutor data. Graded by hand through the full app (German check, app-written answers, reply gate) on 40 new conversations: 11.1% of replies with a mistake. In the app, verdicts, corrections, word meanings and rule-card answers are written by the app itself; the model adds the conversation around them.

Files

• laura-ar-qwen3-4b-Q4_K_M.gguf — Q4_K_M, 2.50 GB