awesome-llm-token-optimization

A curated list of strategies, tools, and papers for cutting LLM token costs in production. It's the playbook behind running Kestrel cheap: the same techniques took its cost per scoring run from $6.51 to $0.84.

• Batching, caching, and provider routing -- the levers that actually move an LLM bill.
• Practitioner-tested, not aggregated: Kestrel runs on this playbook every day.

Where to look:

1. README.md -- the whole list; start at the batching and caching sections.
2. CONTRIBUTING.md -- the inclusion bar that keeps it a guide instead of a link dump.