CLIProxyAPI benchmark · Control and decision

Stop at the current gate

This revision executed no inference; preflight, canaries, and high volume remain blocked.

Primary local sources: BENCHMARK-MEASUREMENT-PLAN.md, model-context-matrix.csv, and planned-run-manifest.json.
BENCHMARK-MEASUREMENT-PLAN.md · model-context-matrix.csv · planned-run-manifest.json

The idea in plain language

The recommended next authorization covers a reversible window, freeze, administrative preflight, safe configuration, up to 90 text canaries plus one image lane, restore, and an eligibility report. The large grid is discussed only after that.

Running example: send the same sealed parcel — nonce, checksum, and three needles — through all six lanes. If the harness changes the wrapping, you are measuring the lane too. The analogy breaks because context, tools, and thinking carry semantics, not size alone.

The future sequence is freeze → preflight → canaries → short grid → long grid → behavior → image → restore → report. unplanned_requests_allowed=0 and concurrency 1 bound scope. Catalog drift, identity mismatch, authentication failure, timeout, concurrent traffic, or an unplanned request stop the run.

Verifiable excerpt

inference_executed_by_revision = false\npreflight_canaries_approved = false\nhigh_volume_benchmark_approved = false\nmax_text_canaries = 15 models × 6 lanes = 90

outputs/BENCHMARK-MEASUREMENT-PLAN.md · outputs/model-context-matrix.csv · outputs/planned-run-manifest.json

The path in one picture

Running example: the same nonce + checksum crosses every lane; the failure belongs to the smallest boundary where it reproduces.

Retrieval check

What is the current operational state?