Local AI Registry

Speed sweep

gemma-4-31b-int4-rtx-3090-24gb-vllm-tp2-sweep

gemma-4-31b-int4-rtx-3090-24gb-vllm-tp2-sweep
JSON APIBack to speed sweep search

Record

Recipe
gemma-4-31b-int4-rtx-3090-24gb-vllm-tp2
Measured
2026-06-27
Points
1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatus
—2,048—135.171historical

Remaining fields

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

id
gemma-4-31b-int4-rtx-3090-24gb-vllm-tp2-sweep
measured at
2026-06-27
recipe id
gemma-4-31b-int4-rtx-3090-24gb-vllm-tp2
schema version
local-ai-registry/v1

metrics

inference engine version
0.26.0
latest point at
2026-06-27
max context tokens
2,048
peak generation tps
135.12
point count
1

rows

concurrencycontext tokensdecode tok sdecode tok s per streamoutput tokenspeak vram gbprefill tok ssamples
Unknown2,048135.12Unknown045.1Unknown1
Provenance & metadata (1)