It generates IELTS Academic practice content across all four sections (Reading, Writing, Listening, Speaking). Full model details, examples, supported types, and limitations are on the main model page: 👉 https://huggingface.co/preparebuddy/ielts-3b

Content generator, not an assessment tool. A specialised fine-tune of SmolLM3-3B — not a from-scratch foundation model. Verdict answers (TFNG/YNNG) are ~70–75% accurate — review before use.

Tip: if LM Studio's MLX runtime gives a load error (a known LM Studio issue), use the GGUF build preparebuddy/ielts-3b-gguf instead — it loads via the rock-solid llama.cpp runtime, on any OS.

Run it in LM Studio (Mac, Apple Silicon)

In LM Studio, search preparebuddy/ielts-3b-mlx and download it.
Load it, and in the system prompt paste:

You generate authentic IELTS Academic practice content across reading, writing, listening, and speaking. Produce passages, transcripts, tasks, questions, and answer keys or model answers as appropriate to the section. Use IELTS-style register: academic, neutral, factually plausible. This is content generation, not assessment.
In the chat box, use the structured prompt, e.g.:
<TEST=IELTS><SECTION=READING><TYPE=TFNG><DIFF=medium><TOPIC=solar power> Generate a short passage with 4 True/False/Not Given statements and an answer key.
Settings: temperature 0.3 for verdict questions (TFNG/YNNG/MCQ), 0.7 for passages/writing/speaking; top_p 0.9.

Run it with mlx-lm (CLI)

bash

pip install mlx-lm
python -m mlx_lm generate --model preparebuddy/ielts-3b-mlx \
  --prompt "<TEST=IELTS><SECTION=WRITING><TYPE=TASK2><DIFF=medium><TOPIC=remote work> Generate an IELTS Academic Writing Task 2 question with word limit and timing." --temp 0.7

(Apply the chat template / system prompt for best results — see the main model page.)

License

Apache-2.0, inheriting from SmolLM3-3B. Built by PrepareBuddy.