Speed sweep

pg-019f2c34-dffe-78f2-94cc-60129e5b27f9-sweep

pg-019f2c34-dffe-78f2-94cc-60129e5b27f9-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-nvidia-nemotron-3-super-120b-a12b-4bit-4bit-apple-m5-max-128gb-mlx-c78122e569
Measured: 2026-07-04T08:18:10.554Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	51.5	—	observed
1	32,832	—	48.2	—	observed
1	258,111	—	32	—	observed
1	32,768	—	—	32,365	observed

Remaining fields

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

id: pg-019f2c34-dffe-78f2-94cc-60129e5b27f9-sweep
measured at: 2026-07-04T08:18:10.554Z
recipe id: pg-mlx-community-nvidia-nemotron-3-super-120b-a12b-4bit-4bit-apple-m5-max-128gb-mlx-c78122e569
schema version: local-ai-registry/v1

metrics

base memory bytes: 69643684812
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 48.165
decode8k context tokens: 8,256
decode8k tps: 51.47
decode max context tokens: 258,111
decode max context tps: 32.024
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-04T08:18:10.554Z
max context tokens: 258,111
max prompt tokens: 258,047
memory8k bytes: 77217455288
memory8k context tokens: 8,256
memory max context bytes: 81244019960
memory max context tokens: 258,111
peak generation tps: 50.336
peak memory bytes: 81244019960
peak prompt tps: 58.961
point count: 223
ttft32k cached prompt tokens: 28,736
ttft32k context tokens: 32,768
ttft32k seconds: 32.365

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	51.47	51.47	Unknown	71.914	Unknown	1
1	32,832	48.165	48.165	Unknown	75.664	Unknown	1
1	258,111	32.024	32.024	Unknown	75.664	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f2c34-dffe-78f2-94cc-60129e5b27f9
repository: local.ai ↗