Speed sweep

pg-019f7582-f247-7798-98d7-75e233563b27-sweep

pg-019f7582-f247-7798-98d7-75e233563b27-sweep

JSON API Back to speed sweep search

Record

Recipe: pg-mlx-community-qwen3-5-9b-mlx-4bit-4bit-apple-m3-ultra-96gb-60c-mlx-6d2c74a819
Measured: 2026-07-18T13:55:43.763Z
Points: 3

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	90.8	—	observed
1	32,832	—	75.9	37,049.7	observed
1	65,600	—	62.2	—	observed

Remaining fields

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

id: pg-019f7582-f247-7798-98d7-75e233563b27-sweep
measured at: 2026-07-18T13:55:43.763Z
recipe id: pg-mlx-community-qwen3-5-9b-mlx-4bit-4bit-apple-m3-ultra-96gb-60c-mlx-6d2c74a819
schema version: local-ai-registry/v1

metrics

base memory bytes: 5500006038
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 75.898
decode8k context tokens: 8,256
decode8k tps: 90.822
decode max context tokens: 65,600
decode max context tps: 62.218
decode mode: non-mtp
inference engine version: 0.31.3
latest point at: 2026-07-18T13:55:43.763Z
max context tokens: 65,600
max prompt tokens: 65,536
memory8k bytes: 7515605662
memory8k context tokens: 8,256
memory max context bytes: 15009942194
memory max context tokens: 65,600
peak generation tps: 75.898
peak memory bytes: 15009942194
peak prompt tps: 884.434
point count: 56
ttft32k cached prompt tokens: 28,736
ttft32k context tokens: 32,768
ttft32k seconds: 37.05

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	90.822	Unknown	Unknown	6.999	Unknown	1
1	32,832	75.898	Unknown	Unknown	13.979	Unknown	1
1	65,600	62.218	Unknown	Unknown	13.979	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:019f7582-f247-7798-98d7-75e233563b27
repository: local.ai ↗