ray-project/llm-numbers
A reference guide of numbers and benchmarks useful for LLM developers to make back-of-the-envelope calculations.
Not currently ranked — collecting fresh signals.
star history
This repository aggregates key numbers and metrics that LLM developers should know for estimation and optimization. It covers token-to-word ratios, cost savings from prompt engineering techniques like appending “Be Concise,” and API pricing data. The resource is maintained by Anyscale and originated from an internal reference document similar to Jeff Dean’s classic performance numbers guide.
Frequently asked
- What is ray-project/llm-numbers?
- A reference guide of numbers and benchmarks useful for LLM developers to make back-of-the-envelope calculations.
- Is llm-numbers open source?
- Yes — ray-project/llm-numbers is an open-source project tracked on heatdrop.
- How popular is llm-numbers?
- ray-project/llm-numbers has 4.3k stars on GitHub.
- Where can I find llm-numbers?
- ray-project/llm-numbers is on GitHub at https://github.com/ray-project/llm-numbers.