Local AI Registry

Speed sweep

qwen3-6-27b-fp8-rtx-3090-24gb-vllm-tp1-sweep

qwen3-6-27b-fp8-rtx-3090-24gb-vllm-tp1-sweep
JSON APIBack to speed sweep search

Record

Recipe
qwen3-6-27b-fp8-rtx-3090-24gb-vllm-tp1
Measured
2026-04-26
Points
1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatus
—2,048—48.5—historical

Remaining fields

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

id
qwen3-6-27b-fp8-rtx-3090-24gb-vllm-tp1-sweep
measured at
2026-04-26
recipe id
qwen3-6-27b-fp8-rtx-3090-24gb-vllm-tp1
schema version
local-ai-registry/v1

metrics

inference engine version
0.19.1
latest point at
2026-04-26
max context tokens
2,048
peak generation tps
48.5
point count
1

rows

concurrencycontext tokensdecode tok sdecode tok s per streamoutput tokenspeak vram gbprefill tok ssamples
Unknown2,04848.5Unknown1,024UnknownUnknown1
Provenance & metadata (1)