Speed sweep

pg-01a011ee-e475-7310-b4c0-33709fa28f59-sweep

pg-01a011ee-e475-7310-b4c0-33709fa28f59-sweep

Record

Recipe: pg-unsloth-qwen3-6-27b-mtp-gguf-q4-k-m-apple-m4-max-128gb-llama-cpp-75508f85cb
Measured: 2026-08-17T22:54:23.857Z
Points: 4

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status
1	8,256	—	22.2	—	observed
1	32,832	—	18.2	—	observed
1	262,080	—	5.1	—	observed
1	32,768	—	—	222,102.6	observed

Remaining fields

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

id: pg-01a011ee-e475-7310-b4c0-33709fa28f59-sweep
measured at: 2026-08-17T22:54:23.857Z
recipe id: pg-unsloth-qwen3-6-27b-mtp-gguf-q4-k-m-apple-m4-max-128gb-llama-cpp-75508f85cb
schema version: local-ai-registry/v1

metrics

base memory bytes: 34532245504
base memory context tokens: 192
concurrency: 1
decode32k context tokens: 32,832
decode32k tps: 18.169
decode8k context tokens: 8,256
decode8k tps: 22.201
decode max context tokens: 262,080
decode max context tps: 5.127
decode mode: non-mtp
inference engine version: built with AppleClang 17.0.0.17000603 for Darwin arm64
latest point at: 2026-08-17T22:54:23.857Z
max context tokens: 262,080
max prompt tokens: 262,016
memory8k bytes: 35704815616
memory8k context tokens: 8,256
memory max context bytes: 40774811648
memory max context tokens: 262,080
peak generation tps: 18.169
peak memory bytes: 40774811648
peak prompt tps: 147.535
point count: 224
ttft32k cached prompt tokens: 28,668
ttft32k context tokens: 32,768
ttft32k seconds: 222.103

rows

concurrency	context tokens	decode tok s	decode tok s per stream	output tokens	peak vram gb	prefill tok s	samples
1	8,256	22.201	22.201	Unknown	33.253	Unknown	1
1	32,832	18.169	18.169	Unknown	37.975	Unknown	1
1	262,080	5.127	5.127	Unknown	37.975	Unknown	1
1	32,768	Unknown	Unknown	Unknown	Unknown	Unknown	1

Provenance & metadata (1)

source

paths: publication:pg-20260827T060320709Z, run:01a011ee-e475-7310-b4c0-33709fa28f59
repository: local.ai ↗