DeepSeek-V4-Flash · UD-IQ2_XXS
This page aggregates 1 real-world runs of DeepSeek-V4-Flash (UD-IQ2_XXS) on 2x NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
19.39 tok/s
Decode
Decode speed
692.41 tok/s
Prefill
Prefill speed
0.12 s
TTFT
Time to first token
51.17 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
GitHub
Source platforms
2026-09-27
Last verified
Performance
- llama.cpp · UD-IQ2_XXS (current)Decode 19.39 · Prefill 692.41 ·
Core figures
- Decode (avg)
- 19.39 tok/s
- Prefill (avg)
- 692.41 tok/s
- TTFT (avg)
- 0.12 s
- VRAM (avg)
- 51.17 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
- UD-IQ2_XXS
- Framework
- llama.cpp
- Version
- b10236
- Context length
- 204800 tokens
- GPU layers
- 99
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- 2x NVIDIA RTX 5090
- Nominal VRAM
- 64 GB
- Measured VRAM (avg)
- 51.17 GB
- OS
- Ubuntu 24.04.4 LTS
- Driver
- 610.43.03
- CUDA
- 13.3
Get started
Weights for this deployment
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
b10236 · Ubuntu 24.04.4 LTS · CUDA 13.3 · 204800 ctx
19.39 tok/s
Decode
692.41 tok/s
Prefill
0.12 s
TTFT
51.17 GB
VRAM
—
MTP
— W
Power draw
2×5090 +9950X3D/123GiB RAM，DeepSeek-V4-Flash-0731 IQ2_XXS，专家CPU卸载+DSpark。narrative均值19.39t/s，短prompt TTFT117ms；9980tok prefill692.41t/s；bench峰值52398MiB≈51.17GiB。415W仅末尾瞬时双卡读数之和，不代表测试平均功耗，已置空。