Expand description
2026-09-26: How a BFCL run is presented (table, summary tiles, gate metrics) and its run verdict, including the MLPerf-edge floors.
Owner: bench, BFCL benchmark. Invariants: none beyond the types.
Constants§
- MLPERF_
FLOOR_ CHECKPOINTS - 2026-09-26: The only served models whose run verdict is gated on the MLPerf
floors (
floor_verdict). The summary tiles style every model against the floors. - MLPERF_
FLOOR_ NORMALIZED - MLPERF_
FLOOR_ OVERALL - 2026-09-26: MLPerf-edge floors for Qwen3.6-27B: the llama.cpp Q4_K_M
reference (86.23 / 87.96, recorded in
kernels/gb10/qwen3.6-27b/BENCH.toml) times 0.97.
Functions§
- is_
mlperf_ submission_ checkpoint - 2026-09-26: Compared ignoring ASCII case, so a serve that spells the org differently still matches.