Using pinstripes with LM Studio

What you'll need

- [LM Studio](https://lmstudio.ai) v0.2.20 or later

- A pinstripes API key — get one at [pinstripes.io/signup](/signup)

Step 1 — Open the Developer tab

In LM Studio, click the **<>** (Developer) icon in the left sidebar.

Step 2 — Connect to a remote server

Switch from **Local Server** to **Remote Server** at the top of the panel. Enter:

Click **Connect**.

Step 3 — Select a model

Open the model picker dropdown. pinstripes' models will appear — select ps/qwen3.6-a3b or any other model in your plan.

Step 4 — Start chatting

Switch to the **Chat** tab. All inference now routes through pinstripes' API. Your prompts never touch your GPU.

Why bother?

LM Studio's local inference is great for experimentation, but it serialises requests, so you can only handle one conversation at a time. pinstripes routes to a fleet of workers, so you can open multiple LM Studio windows simultaneously at full throughput without saturating your machine.

Troubleshooting

**"Connection refused"**: Make sure you're using https://api.pinstripes.io/v1 (with the /v1 suffix) as the base URL.

**"401 Unauthorized"**: Double-check your API key starts with sk-ps-. Keys created before June 2026 may need to be regenerated in the [dashboard](/dashboard).

**High latency on first token**: The first request after a period of inactivity may be slower as workers spin up. Subsequent requests are fast.