Speed sweep

pg-019f2c34-cff5-7d88-893d-5a7ed6218d18-sweep

pg-019f2c34-cff5-7d88-893d-5a7ed6218d18-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-nvidia-nemotron-3-nano-30b-a3b-nvfp4-nvfp4-apple-m3-ultra-96gb-60c-mlx-7e0d994571
Measured: 2026-07-04T08:18:06.448Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	98.3	—	observed
1	32,832	—	89.3	—	observed
1	258,111	—	48	—	observed
1	32,768	—	—	22,434.8	observed

Remaining fields

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

id: pg-019f2c34-cff5-7d88-893d-5a7ed6218d18-sweep
measured at: 2026-07-04T08:18:06.448Z
recipe id: pg-mlx-community-nvidia-nemotron-3-nano-30b-a3b-nvfp4-nvfp4-apple-m3-ultra-96gb-60c-mlx-7e0d994571
schema version: local-ai-registry/v1

metrics

base memory bytes: 19924908528
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 89.315
decode8k context tokens: 8,256
decode8k tps: 98.282
decode max context tokens: 258,111
decode max context tps: 48
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-04T08:18:06.448Z
max context tokens: 258,111
max prompt tokens: 258,047
memory8k bytes: 22882262138
memory8k context tokens: 8,256
memory max context bytes: 26110334086
memory max context tokens: 258,111
peak generation tps: 89.315
peak memory bytes: 26110334086
peak prompt tps: 1,460.587
point count: 223
ttft32k cached prompt tokens: 28,736
ttft32k context tokens: 32,768
ttft32k seconds: 22.435

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	98.282	98.282	Unknown	21.311	Unknown	1
1	32,832	89.315	89.315	Unknown	24.317	Unknown	1
1	258,111	48	48	Unknown	24.317	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f2c34-cff5-7d88-893d-5a7ed6218d18
repository: local.ai ↗