1 GPU on the selected test

embeddinggemma-2

GPUs are compared only on one exact test of this model. Change the test to see other conditions.

Run this model

Choose your operating system and enter a model repository. Launch support is checked separately from performance results.

Test conditions

Every GPU below ran exactly these conditions.
Task
Embedding · Text → Vectors
Workload
Text → Vectors · embedding-sanity-v1 · Test duration 6 s
Batch size
Batch 4
Format · Engine
bf16 · sentence-transformers · PyTorch 2.13.0+cu130
Unit
vectors/sec

Occupancy includes engine buffers and reserved KV cache. It is not the minimum VRAM required by the model.

Reports combined counts reports, not inference requests. Request successes and failures are listed in each report.

Protocol details

native-forward-v1 · embedding-sanity-v1 · embedding

GPUs on this test

Only one GPU configuration has matching measurements; a cross-hardware comparison is not yet available.

  1. vGPU-32GB · 32 GBPyTorch 2.13.0+cu130
    Mean throughput
    139vectors/sec
    Observed GPU occupancy
    0.8GiB
    P95 latency
    33 ms
    Reports combined
    4
3 more GPUs ran this model under other tests
Sortable table
Hardware ranking · 4 rows
HardwareMean throughput (vectors/sec)P95 (s)TTFT (ms)Observed occupancy (GiB)GPU power (W)Throughput / WVRAM per GPU (GB)Reports combined
#1vGPU-32GB · 32 GBPyTorch 2.13.0+cu130Runs, parameters and setup ↗138.640.033—0.8254.242.56324
L4 · 24 GBNo matching measurements——————24—
RTX PRO 6000 Blackwell Server Edition · 96 GBNo matching measurements——————96—
T4 · 16 GBNo matching measurements——————16—

Occupancy includes engine buffers and reserved KV cache. It is not the minimum VRAM required by the model.

Reports combined counts reports, not inference requests. Request successes and failures are listed in each report.

How results are grouped

Results merge successful runs of the same model, precision and workload. CPU thread counts and CUDA wheel builds may differ; actual PyTorch builds are shown. Other recorded protocol settings remain separate.

Updated 11/5/2026, 11:02:27 PM