About AI Latency Tracker
AI Latency Tracker (llmlatency.dev) measures the latency and uptime of AI inference APIs from several regions and publishes the results as open data. It started on , the day of the first measurement.
Who runs it
Founded and run by Maksym Babenko, a software engineer based in Prague, Czechia. He builds and operates the probes, the data pipeline and this site. LinkedIn · GitHub · Personal site.
What is measured
46 providers are probed from 4 regions (Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)). The database currently holds 3,987,912 measurements, collected since 2026-07-23; raw rows are kept for 90 days. Probes call each API directly, not through a gateway. How it works: methodology.
Independence and funding
- Self-funded: no investors and no sponsorship from any AI provider.
- No provider pays for its position. Optional paid data pilots cannot change ranking order (data & licensing).
- The founder is not employed by any provider measured on this site.
Open data and code
- Source code: https://github.com/mazamaka/llm-latency-tracker
- https://doi.org/10.5281/zenodo.22764636
- https://huggingface.co/datasets/llmlatency/llm-latency-tracker
- JSON API · OpenAPI · MCP endpoint · llms.txt — free under CC BY 4.0
Contact
contact@llmlatency.dev · company page on LinkedIn. Think a number is wrong? How corrections work.