Speed sweep

pg-019f44a9-a340-7685-96f5-5b1bb1a39ecf-sweep

pg-019f44a9-a340-7685-96f5-5b1bb1a39ecf-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-gpt-oss-20b-mxfp4-q8-mxfp4-q8-apple-m4-max-128gb-mlx-be1153bfb6
Measured: 2026-07-09T02:16:35.896Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	2,111	—	101.2	—	observed
1	32,831	—	68.3	—	observed
1	131,007	—	35.2	—	observed
1	32,767	—	—	28,192.5	observed

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

id: pg-019f44a9-a340-7685-96f5-5b1bb1a39ecf-sweep
measured at: 2026-07-09T02:16:35.896Z
recipe id: pg-mlx-community-gpt-oss-20b-mxfp4-q8-mxfp4-q8-apple-m4-max-128gb-mlx-be1153bfb6
schema version: local-ai-registry/v1

metrics

base memory bytes: 12284006990
base memory context tokens: 191
concurrency: 1
decode32k context tokens: 32,831
decode32k tps: 68.335
decode8k context tokens: 2,111
decode8k tps: 101.215
decode max context tokens: 131,007
decode max context tps: 35.231
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-09T02:16:35.896Z
max context tokens: 131,007
max prompt tokens: 130,943
memory8k bytes: 13005211082
memory8k context tokens: 2,111
memory max context bytes: 18596145006
memory max context tokens: 131,007
peak generation tps: 68.335
peak memory bytes: 18596145006
peak prompt tps: 1,162.259
point count: 16
ttft32k cached prompt tokens: 0
ttft32k context tokens: 32,767
ttft32k seconds: 28.193

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	2,111	101.215	101.215	Unknown	12.112	Unknown	1
1	32,831	68.335	68.335	Unknown	17.319	Unknown	1
1	131,007	35.231	35.231	Unknown	17.319	Unknown	1
1	32,767	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f44a9-a340-7685-96f5-5b1bb1a39ecf
repository: local.ai ↗