Skip to content

Benchmark: fused profile vs single-model baseline #14

Description

@stubbi

Wire a benchmark that runs a fusion profile against a single-model baseline and records quality, cost (tokens), and latency. Per the no-unverified-claims rule, this is the source for any number in the README.

Acceptance: a committed, re-runnable benchmark that emits a quality/cost/latency table.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:benchbenchmarks and evidence

    Type

    No type

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions