Back to search results
Qwen3.5-4B · IQ2_M
AMD Radeon RX 9070 XT16GBllama.cpp v0.42L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (IQ2_M) on AMD Radeon RX 9070 XT with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
196.42 tok/s
Decode
Decode speed
767.31 tok/s
Prefill
Prefill speed
1.36 s
TTFT
Time to first token
6.47 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ2_M (current)Decode 196.42 · Prefill 767.31 ·
- Ollama · FP16 Decode 69.03 · Prefill 237.27 ·
- Ollama · Q6_K Decode 103.08 · Prefill 389.31 ·
- Ollama · Q3_K_L Decode 133.65 · Prefill 600.15 ·
Core figures
- Decode (avg)
- 196.42 tok/s
- Prefill (avg)
- 767.31 tok/s
- TTFT (avg)
- 1.36 s
- VRAM (avg)
- 6.47 GB
- MTP acceptance rate
- —
- TTFB
- 1.57 GB
- Power draw
- 229.2 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- IQ2_M
- Framework
- llama.cpp
- Version
- v0.42
- Context length
- 4096 tokens
- Batch size
- 512
- GPU layers
- 20
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- AMD Radeon RX 9070 XT
- Nominal VRAM
- 16 GB
- Measured VRAM (avg)
- 6.47 GB
- OS
- Windows 11 23H2
- Driver
- amdgpu 6.12
- CUDA
- 12.4
- Power draw
- 229.2 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.42 · Windows 11 23H2 · CUDA 12.4 · 4096 ctx
196.42 tok/s
Decode
767.31 tok/s
Prefill
1.36 s
TTFT
6.47 GB
VRAM
—
MTP
229.2 W
Power draw
L2 batch seed 2026-10-18: llama.cpp on hw_id=16 running model_id=48 at IQ2_M