LLM Access for Students
For course projects, a small OpenRouter budget is usually allocated to each project, but you can also work with LLMs for free. Here is where to get access, and how to keep your costs low.
Where to access LLMs
One API for many cloud models, or run models locally at no cost.
-
Recommended · free models
Recommended for cloud LLM access. One API for many commercial and open models, so you can switch models without changing your application. It also provides free models for development and experimentation.
-
Local · no API cost
Run models such as Llama, Qwen, Gemma, and others locally on your computer, with no API cost. Particularly useful for development and testing.
Keep your LLM costs low
API costs can grow quickly when processing large datasets. A few simple rules help.
- Debug on a small sample first. Do not run 10,000 examples before confirming that your pipeline works correctly on 10–50.
- Use cheaper or free models during development, and use more expensive models only when necessary.
- Batch requests when processing large datasets. OpenRouter provides a Batch API for submitting many inference requests together.
- Use prompt caching when many requests share the same long instructions or context. See the OpenRouter Prompt Caching guide.
- Keep prompts and outputs as short as practical. You normally pay for both input and output tokens.
- Monitor usage and cost while developing, rather than discovering the cost after a large experiment.
Rule of thumb: develop and debug small and cheap, then scale up only after everything works.