Speed sweep

pg-019f7582-e781-70b6-8c2b-683941af8847-sweep

pg-019f7582-e781-70b6-8c2b-683941af8847-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-gemma-4-12b-coder-fable5-composer2-5-8bit-8bit-apple-m3-ultra-96gb-80c-mlx-6a59710e45
Measured: 2026-07-18T13:55:40.981Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	39.4	—	observed
1	32,832	—	37.4	—	observed
1	65,600	—	35.9	—	observed
1	32,768	—	—	322,725.8	observed

Remaining fields

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

id: pg-019f7582-e781-70b6-8c2b-683941af8847-sweep
measured at: 2026-07-18T13:55:40.981Z
recipe id: pg-mlx-community-gemma-4-12b-coder-fable5-composer2-5-8bit-8bit-apple-m3-ultra-96gb-80c-mlx-6a59710e45
schema version: local-ai-registry/v1

metrics

base memory bytes: 12987425702
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 37.401
decode8k context tokens: 8,256
decode8k tps: 39.411
decode max context tokens: 65,600
decode max context tps: 35.861
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-18T13:55:40.981Z
max context tokens: 65,600
max prompt tokens: 65,536
memory8k bytes: 15420784768
memory8k context tokens: 8,256
memory max context bytes: 21220655222
memory max context tokens: 65,600
peak generation tps: 37.132
peak memory bytes: 21220655222
peak prompt tps: 7.586
point count: 56
ttft32k cached prompt tokens: 0
ttft32k context tokens: 32,768
ttft32k seconds: 322.726

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	39.411	39.411	Unknown	14.362	Unknown	1
1	32,832	37.401	37.401	Unknown	19.763	Unknown	1
1	65,600	35.861	35.861	Unknown	19.763	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f7582-e781-70b6-8c2b-683941af8847
repository: local.ai ↗