Azimuth Bench

Rows

lane=all

Integrity

PASS

fail-closed schema gate

Schema

1.0.0

canonical bundle

Run pages

artifact detail pages

Chart

Structured throughput leaderboard

Top rows

Model	Lane	Think	JSON tok/s	First answer ms
Phi 4 Mini	core	default	146.3	95.1
Gemma 3 4B	core	default	134.4	137.7
Qwen2.5 Coder 7B	core	default	107.3	72.1
Qwen2.5 Instruct 7B	core	default	107.2	80.0
DeepSeek R1 Distill Llama 8B	core	default	103.9	93.4

Tradeoff

Evidence

report/data/run.json
run metadata and provenance status
report/data/summary.json
canonical leaderboard rows
report/data/compare.json
scoped compare projection with blocked comparisons
report/data/runs/<artifact_key>/
per-run run/summary/machine/provider/model/cases bundle
report/data/site_manifest.json
hosted site route contract
report/exports/share_leaderboard.svg
deterministic share snapshot (top rows)
report/exports/share_compare.svg
deterministic scoped compare card