DeepSeek-V4-Flash · IQ2_XS
This page aggregates 1 real-world runs of DeepSeek-V4-Flash (IQ2_XS) on NVIDIA RTX 5070 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
5.5 tok/s
Decode
Decode speed
41.3 tok/s
Prefill
Prefill speed
2.42 s
TTFT
Time to first token
12 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ2_XS (current)Decode 5.5 · Prefill 41.3 ·
Core figures
- Decode (avg)
- 5.5 tok/s
- Prefill (avg)
- 41.3 tok/s
- TTFT (avg)
- 2.42 s
- VRAM (avg)
- 12 GB
- MTP acceptance rate
- —
- TTFB
- — GB
- Power draw
- — W
Configuration
Member-level fields are taken from the most recent run
- Model
- DeepSeek-V4-Flash
- Quantization
- IQ2_XS
- 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 5070
- Nominal VRAM
- 12 GB
- Measured VRAM (avg)
- 12 GB
- OS
- Linux
- Driver
- 570.65
- CUDA
- 12.6
Get started
Original model resources
Base model resources may not match this quantization. Use the verified deployment weights above when available.
Sources & evidence
1 measured records in total, each traceable to its original source
b4200 · Linux · CUDA 12.6 · 4096 ctx
5.5 tok/s
Decode
41.3 tok/s
Prefill
2.42 s
TTFT
12 GB
VRAM
—
MTP
— W
Power draw
L0 理论估算：基于显存带宽 ÷ 模型体积公式，结合框架实测效率因子校准