Back to search results
Ministral-3-8B-Instruct-2512 · Q3_K_S
NVIDIA RTX 509032GBvLLM v0.41L2 Cross-framework verified
This page aggregates 1 real-world runs of Ministral-3-8B-Instruct-2512 (Q3_K_S) on NVIDIA RTX 5090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
337.39 tok/s
Decode
Decode speed
1,195.07 tok/s
Prefill
Prefill speed
0.67 s
TTFT
Time to first token
13.02 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
11 days ago
Last verified
Performance
- llama.cpp · IQ1_S Decode 381.92 · Prefill 1480.95 ·
- vLLM · Q3_K_S (current)Decode 337.39 · Prefill 1195.07 ·
Core figures
- Decode (avg)
- 337.39 tok/s
- Prefill (avg)
- 1,195.07 tok/s
- TTFT (avg)
- 0.67 s
- VRAM (avg)
- 13.02 GB
- MTP acceptance rate
- —
- TTFB
- 2.7 GB
- Power draw
- 399.6 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Ministral-3-8B-Instruct-2512
- Quantization
- Q3_K_S
- Framework
- vLLM
- Version
- v0.41
- Context length
- 65536 tokens
- Batch size
- 8
- GPU layers
- 999
- Flash Attention
- Off
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)
- 13.02 GB
- OS
- Ubuntu 22.04
- Driver
- 565.57
- CUDA
- 12.6
- Power draw
- 399.6 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.41 · Ubuntu 22.04 · CUDA 12.6 · 65536 ctx
337.39 tok/s
Decode
1,195.07 tok/s
Prefill
0.67 s
TTFT
13.02 GB
VRAM
—
MTP
399.6 W
Power draw
L2 batch seed 2026-10-24: vLLM on hw_id=13 running model_id=74 at Q3_K_S