Speed sweep

nvidia-nemotron-labs-3-puzzle-75b-a9b-bf16-nvfp4-dgx-spark-gb10-128gb-vllm-tp1-sweep

nvidia-nemotron-labs-3-puzzle-75b-a9b-bf16-nvfp4-dgx-spark-gb10-128gb-vllm-tp1-sweep

Record

Recipe
nvidia-nemotron-labs-3-puzzle-75b-a9b-bf16-nvfp4-dgx-spark-gb10-128gb-vllm-tp1
Measured
2026-07-08T02:59:57.744Z
Points
1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatus
262,144231.6observed

Remaining fields

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

id
nvidia-nemotron-labs-3-puzzle-75b-a9b-bf16-nvfp4-dgx-spark-gb10-128gb-vllm-tp1-sweep
measured at
2026-07-08T02:59:57.744Z
recipe id
nvidia-nemotron-labs-3-puzzle-75b-a9b-bf16-nvfp4-dgx-spark-gb10-128gb-vllm-tp1
schema version
local-ai-registry/v1

metrics

inference engine version
0.24.0+aeon.sm121a.dflash
latest point at
2026-07-08T02:59:57.744Z
max context tokens
262,144
peak generation tps
231.55
point count
1

rows

concurrencycontext tokensdecode tok sdecode tok s per streamoutput tokenspeak vram gbprefill tok ssamples
Unknown262,144231.55Unknown512109.81Unknown1
Provenance & metadata (1)