LargestLLM (baseline)

LargestLLM is a baseline router: it always routes to the largest candidate model (based on the size field in llm_data).

Notebook: https://github.com/ulab-uiuc/LLMRouter/blob/main/notebooks/largest_llm/01_largest_llm_inference.ipynb

Router docs: https://github.com/ulab-uiuc/LLMRouter/blob/main/llmrouter/models/largest_llm/README.md

Config

• Test config: https://github.com/ulab-uiuc/LLMRouter/blob/main/configs/model_config_test/largest_llm.yaml

Warning

The notebook currently references configs/model_config_train/largest_llm.yaml, but the repo ships the baseline config under configs/model_config_test/. Use the test config above (or change the notebook path).

Run (CLI)

Route-only (no API calls):

llmrouter infer --router largest_llm --config configs/model_config_test/largest_llm.yaml --query "Explain transformers." --route-only

Full inference (routes + calls the selected model): - Ensure llm_data contains api_endpoint and model for the routed model. - Set API_KEYS (see Installation).

llmrouter infer --router largest_llm --config configs/model_config_test/largest_llm.yaml --query "Explain transformers."

Next

• Compare with SmallestLLM.
• See the router table: Routers.