Speed sweep

pg-01a026ab-7ce6-7e38-99c1-85cd6d8a7ea6-sweep

pg-01a026ab-7ce6-7e38-99c1-85cd6d8a7ea6-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-llama-3-3-70b-instruct-q4km-unknown-apple-m5-pro-64gb-llama-cpp-489fec45df
Measured: 2026-08-21T23:32:47.961Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	6.2	—	observed
1	32,832	—	4.9	—	observed
1	40,896	—	4.6	—	observed
1	32,768	—	—	739,084.4	observed

Remaining fields

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

id: pg-01a026ab-7ce6-7e38-99c1-85cd6d8a7ea6-sweep
measured at: 2026-08-21T23:32:47.961Z
recipe id: pg-llama-3-3-70b-instruct-q4km-unknown-apple-m5-pro-64gb-llama-cpp-489fec45df
schema version: local-ai-registry/v1

metrics

base memory bytes: 55526342656
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 4.903
decode8k context tokens: 8,256
decode8k tps: 6.217
decode max context tokens: 40,896
decode max context tps: 4.585
decode mode: non-mtp
inference engine version: unknown
latest point at: 2026-08-21T23:32:47.961Z
max context tokens: 40,896
max prompt tokens: 40,832
memory8k bytes: 55619158016
memory8k context tokens: 8,256
memory max context bytes: 55759044608
memory max context tokens: 40,896
peak generation tps: 4.903
peak memory bytes: 55759044608
peak prompt tps: 44.336
point count: 35
ttft32k cached prompt tokens: 28,735
ttft32k context tokens: 32,768
ttft32k seconds: 739.084

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	6.217	6.217	Unknown	51.799	Unknown	1
1	32,832	4.903	4.903	Unknown	51.93	Unknown	1
1	40,896	4.585	4.585	Unknown	51.93	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:01a026ab-7ce6-7e38-99c1-85cd6d8a7ea6
repository: local.ai ↗