Speed sweep

pg-019f3295-e56f-7bbc-80e1-a80cae3f68f2-sweep

pg-019f3295-e56f-7bbc-80e1-a80cae3f68f2-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-gpt-oss-120b-mxfp4-q8-mxfp4-q8-apple-m3-ultra-96gb-80c-mlx-ed712f6469
Measured: 2026-07-05T14:01:52.235Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,255	—	66.8	—	observed
1	32,831	—	53.8	—	observed
1	127,039	—	30.5	—	observed
1	32,767	—	—	24,831.8	observed

Remaining fields

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

id: pg-019f3295-e56f-7bbc-80e1-a80cae3f68f2-sweep
measured at: 2026-07-05T14:01:52.235Z
recipe id: pg-mlx-community-gpt-oss-120b-mxfp4-q8-mxfp4-q8-apple-m3-ultra-96gb-80c-mlx-ed712f6469
schema version: local-ai-registry/v1

metrics

base memory bytes: 63553160102
base memory context tokens: 191
concurrency: 1
decode32k context tokens: 32,831
decode32k tps: 53.818
decode8k context tokens: 8,255
decode8k tps: 66.829
decode max context tokens: 127,039
decode max context tps: 30.475
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-05T14:01:52.235Z
max context tokens: 127,039
max prompt tokens: 126,975
memory8k bytes: 64824624762
memory8k context tokens: 8,255
memory max context bytes: 73115584934
memory max context tokens: 127,039
peak generation tps: 53.818
peak memory bytes: 73115584934
peak prompt tps: 1,319.556
point count: 16
ttft32k cached prompt tokens: 0
ttft32k context tokens: 32,767
ttft32k seconds: 24.832

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,255	66.829	66.829	Unknown	60.373	Unknown	1
1	32,831	53.818	53.818	Unknown	68.094	Unknown	1
1	127,039	30.475	30.475	Unknown	68.094	Unknown	1
1	32,767	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f3295-e56f-7bbc-80e1-a80cae3f68f2
repository: local.ai ↗