Back to search results
gemma-4-12B-it · Q4_K_M
NVIDIA RTX 509032GBllama.cpp b4200L0 Self-reported
This page aggregates 1 real-world runs of gemma-4-12B-it (Q4_K_M) on NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
191.5 tok/s
Decode
Decode speed
1,436.3 tok/s
Prefill
Prefill speed
0.07 s
TTFT
Time to first token
7.8 GB
VRAM
VRAM usage
L0 Self-reported
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- vLLM · Q8_0 Decode 117.2 · Prefill 879 ·
- vLLM · Q5_K_M Decode 171.1 · Prefill 1283.3 ·
- vLLM · Q4_K_M Decode 208.7 · Prefill 1565.3 ·
- llama.cpp · Q8_0 Decode 107.5 · Prefill 806.3 ·
- llama.cpp · Q5_K_M Decode 157.1 · Prefill 1178.3 ·
- llama.cpp · Q4_K_M (current)Decode 191.5 · Prefill 1436.3 ·
Core figures
- Decode (avg)
- 191.5 tok/s
- Prefill (avg)
- 1,436.3 tok/s
- TTFT (avg)
- 0.07 s
- VRAM (avg)
- 7.8 GB
- MTP acceptance rate
- —
- TTFB
- — GB
- Power draw
- — W
Configuration
Member-level fields are taken from the most recent run
- Model
- gemma-4-12B-it
- Quantization
- Q4_K_M
- Framework
- llama.cpp
- Version
- b4200
- Context length
- 4096 tokens
- Batch size
- 512
- GPU layers
- 99
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 5090
- Nominal VRAM
- 32 GB
- Measured VRAM (avg)
- 7.8 GB
- OS
- Linux
- Driver
- 570.65
- CUDA
- 12.6
Sources & evidence
1 measured records in total, each traceable to its original source
b4200 · Linux · CUDA 12.6 · 4096 ctx
191.5 tok/s
Decode
1,436.3 tok/s
Prefill
0.07 s
TTFT
7.8 GB
VRAM
—
MTP
— W
Power draw
L0 理论估算（add-1000-records-batch3）