Speed sweep

pg-01a026ab-4e24-7d9c-9f0b-689b72810392-sweep

pg-01a026ab-4e24-7d9c-9f0b-689b72810392-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-llama-3-3-70b-instruct-q4km-unknown-apple-m3-ultra-96gb-80c-llama-cpp-31d3f23f58
Measured: 2026-08-21T23:32:35.989Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	13.5	—	observed
1	32,832	—	11.1	—	observed
1	131,008	—	6.3	—	observed
1	32,768	—	—	360,261.9	observed

Remaining fields

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

id: pg-01a026ab-4e24-7d9c-9f0b-689b72810392-sweep
measured at: 2026-08-21T23:32:35.989Z
recipe id: pg-llama-3-3-70b-instruct-q4km-unknown-apple-m3-ultra-96gb-80c-llama-cpp-31d3f23f58
schema version: local-ai-registry/v1

metrics

base memory bytes: 85695954944
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 11.125
decode8k context tokens: 8,256
decode8k tps: 13.518
decode max context tokens: 131,008
decode max context tps: 6.328
decode mode: non-mtp
inference engine version: unknown
latest point at: 2026-08-21T23:32:35.989Z
max context tokens: 131,008
max prompt tokens: 130,944
memory8k bytes: 85789687808
memory8k context tokens: 8,256
memory max context bytes: 86333161472
memory max context tokens: 131,008
peak generation tps: 11.125
peak memory bytes: 86333161472
peak prompt tps: 90.956
point count: 112
ttft32k cached prompt tokens: 28,735
ttft32k context tokens: 32,768
ttft32k seconds: 360.262

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	13.518	13.518	Unknown	79.898	Unknown	1
1	32,832	11.125	11.125	Unknown	80.404	Unknown	1
1	131,008	6.328	6.328	Unknown	80.404	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:01a026ab-4e24-7d9c-9f0b-689b72810392
repository: local.ai ↗