Qwen3.8-27B · IQ3_XXS
This page aggregates 1 real-world runs of Qwen3.8-27B (IQ3_XXS) on NVIDIA RTX 4080 Super with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
60.4 tok/s
Decode
Decode speed
453 tok/s
Prefill
Prefill speed
0.22 s
TTFT
Time to first token
10 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- Strata · Q5_K_S Decode 40.4 · Prefill 303 ·
- Strata · IQ3_S Decode 67.5 · Prefill 506.3 ·
- Strata · IQ4_XS Decode 50.4 · Prefill 378 ·
- llama.cpp · IQ4_XS Decode 39.3 · Prefill 294.8 ·
- llama.cpp · Q5_K_S Decode 31.5 · Prefill 236.3 ·
- llama.cpp · IQ3_XXS (current)Decode 60.4 · Prefill 453 ·
- llama.cpp · IQ3_S Decode 52.7 · Prefill 395.3 ·
- llama.cpp · Q6_K Decode 24.2 · Prefill 181.5 ·
Core figures
- Decode (avg)
- 60.4 tok/s
- Prefill (avg)
- 453 tok/s
- TTFT (avg)
- 0.22 s
- VRAM (avg)
- 10 GB
- MTP acceptance rate
- —
- TTFB
- — GB
- Power draw
- — W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.8-27B
- Quantization
- IQ3_XXS
- 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 4080 Super
- Nominal VRAM
- 16 GB
- Measured VRAM (avg)
- 10 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
60.4 tok/s
Decode
453 tok/s
Prefill
0.22 s
TTFT
10 GB
VRAM
—
MTP
— W
Power draw
L0 理论估算（add-1000-records-batch3）