Skip to main content

Deterministic Execution

EigenAI provides verifiable AI inference through deterministic execution using GPUs. See deterministic execution in action at deterministicinference.com.

The EigenAI API is deterministic, meaning that when you send the same request (including identical prompt, parameters, and configuration) multiple times, it produces exactly the same output bit-for-bit each time. This differs from APIs such as OpenAI or Anthropic, which may return slightly different responses for identical inputs because their models are non-deterministic by design.

Deterministic execution enables:

  • Reproducible results. The same request results in the same output. Reproducible results enable reliable workflows with consistent outputs.
  • Verifiability. A EigenAI user can repeat a request made by an application and verify they get the same result.
  • Consistent tool call planning.
  • Simplified debugging of workflows that include AI inference due to consistent AI outputs.