Back to search results
gemma-4-31B-it · IQ2_M
NVIDIA RTX 509032GBllama.cpp v0.48L2 Cross-framework verified
This page aggregates 1 real-world runs of gemma-4-31B-it (IQ2_M) on NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
159.29 tok/s
Decode
Decode speed
615.31 tok/s
Prefill
Prefill speed
0.11 s
TTFT
Time to first token
30.4 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ2_M (current)Decode 159.29 · Prefill 615.31 ·
- Ollama · IQ1_S Decode 146.23 · Prefill 719.62 ·
- vLLM · IQ2_XS Decode 145.1 · Prefill 1088.3 ·
- vLLM · Q5_K_M Decode 66.2 · Prefill 496.5 ·
- vLLM · Q4_K_M Decode 81 · Prefill 607.5 ·
- llama.cpp · IQ2_XS Decode 133.1 · Prefill 998.3 ·
- llama.cpp · Q5_K_M Decode 60.8 · Prefill 456 ·
- llama.cpp · Q4_K_M Decode 74.3 · Prefill 557.3 ·
Core figures
- Decode (avg)
- 159.29 tok/s
- Prefill (avg)
- 615.31 tok/s
- TTFT (avg)
- 0.11 s
- VRAM (avg)
- 30.4 GB
- MTP acceptance rate
- —
- TTFB
- 1.27 GB
- Power draw
- 496.6 W
Configuration
Member-level fields are taken from the most recent run
- Model
- gemma-4-31B-it
- Quantization
- IQ2_M
- Framework
- llama.cpp
- Version
- v0.48
- Context length
- 32768 tokens
- Batch size
- 256
- GPU layers
- 33
- 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)
- 30.4 GB
- OS
- Ubuntu 22.04
- Driver
- 555.85
- CUDA
- 12.4
- Power draw
- 496.6 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.48 · Ubuntu 22.04 · CUDA 12.4 · 32768 ctx
159.29 tok/s
Decode
615.31 tok/s
Prefill
0.11 s
TTFT
30.4 GB
VRAM
—
MTP
496.6 W
Power draw
L2 batch seed 2026-10-10: llama.cpp on hw_id=13 running model_id=63 at IQ2_M