LM Studio
LM Studio gives you more controls than the Ollama app: a system prompt, temperature, presets and a built-in way to chat with documents. A system prompt is a set of instructions that gives the model its role and rules for the whole conversation. Temperature controls how random the output is.
Get the model and chat
Open Discover
Click Discover in the left bar.
Paste the Hugging Face URL
Paste https://huggingface.co/unsloth/Apertus-8B-Instruct-2509-GGUF into the search field. If your computer has 8 GB of memory, paste https://huggingface.co/jondale/Apertus-v1.1-4B-Instruct-GGUF instead and choose its Q4_K_M file (2.43 GB).
Choose Q4_K_M
Choose the Q4_K_M file (5.06 GB). The label describes the quantization: the model’s numbers are stored at lower precision, which makes the file smaller.
Download
Click Download and wait for it to finish.
Load with context length 8192
Load the model and set the context length to 8192 (4096 for the 4B) before you confirm. The context length is how much text the model can hold at once, prompt and answer together. You can only set it when the model loads.
Chat
Type a message in the chat box and send it.
Open the sidebar
Switch to Developer mode to see the sidebar. There you can set a system prompt, change the temperature and save as preset, which stores these settings under a name so you can reuse them later.
To learn what these settings do, see Temperature.
Chat with documents
Attach a PDF
Attach a PDF to your message in a chat.
Ask a question about it
Ask something the document can answer. For example, ask for a summary of one section.
LM Studio handles an attached document in one of two ways:
- Full document: if the whole file fits in the context, LM Studio puts the full text into the prompt, unchanged.
- Retrieval (RAG): if it does not fit, LM Studio selects the passages that seem relevant to your question and adds only those.
Here you get a preview of the RAG you build in Part 2, without writing any code. RAG stands for retrieval-augmented generation: the program fetches the relevant passages first and gives them to the model with your question. The model sees only the text that was put into its context. If the context is too small, or if retrieval picks the wrong passages, the answer gets worse.
Models and sizes are covered in Download the models. Next, try the first prompts.