Arao AI
Back to search results

DeepSeek-R1-Distill-Qwen-14B · IQ2_M

Start a discussion
NVIDIA RTX 409024GBvLLM v0.51L2 Cross-framework verified

This page aggregates 1 real-world runs of DeepSeek-R1-Distill-Qwen-14B (IQ2_M) on NVIDIA RTX 4090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

163.66 tok/s

Decode

Decode speed

743.62 tok/s

Prefill

Prefill speed

2.42 s

TTFT

Time to first token

17.39 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · IQ2_M (current)Decode 163.66 · Prefill 743.62 ·
  2. Ollama · Q3_K_L Decode 117.29 · Prefill 610.83 ·
  3. vLLM · Q5_K_M Decode 82.4 · Prefill 618 ·
  4. vLLM · Q4_K_M Decode 100.8 · Prefill 756 ·
  5. llama.cpp · Q5_K_M Decode 75.6 · Prefill 567 ·
  6. llama.cpp · Q4_K_M Decode 92.5 · Prefill 693.8 ·

Core figures

Decode (avg)
163.66 tok/s
Prefill (avg)
743.62 tok/s
TTFT (avg)
2.42 s
VRAM (avg)
17.39 GB
MTP acceptance rate
—
TTFB
1.31 GB
Power draw
390.2 W

Configuration

Member-level fields are taken from the most recent run

Model
DeepSeek-R1-Distill-Qwen-14B
Quantization
IQ2_M
Framework
vLLM
Version
v0.51
Context length
4096 tokens
Batch size
16
GPU layers
999
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA RTX 4090
Nominal VRAM
24 GB
Measured VRAM (avg)
17.39 GB
OS
Windows 11 24H2
Driver
570.86
CUDA
12.8
Power draw
390.2 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.51 · Windows 11 24H2 · CUDA 12.8 · 4096 ctx

    163.66 tok/s

    Decode

    743.62 tok/s

    Prefill

    2.42 s

    TTFT

    17.39 GB

    VRAM

    —

    MTP

    390.2 W

    Power draw

    L2 batch seed 2026-10-18: vLLM on hw_id=1 running model_id=71 at IQ2_M