Back to search results
Qwen3-4B-Thinking-2507 · IQ2_M
AMD Ryzen AI Max+ 395128GBllama.cpp v0.52L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (IQ2_M) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
178.19 tok/s
Decode
Decode speed
583.74 tok/s
Prefill
Prefill speed
1.53 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
5 days ago
Last verified
Performance
- llama.cpp · Q5_K_M Decode 107.36 · Prefill 478 ·
- llama.cpp · Q8_0 Decode 78.29 · Prefill 263.52 ·
- Ollama · Q3_K_S Decode 145.01 · Prefill 581.64 ·
- llama.cpp · IQ2_M (current)Decode 178.19 · Prefill 583.74 ·
- Strata · Q3_K_L Decode 190.53 · Prefill 851.53 ·
Core figures
- Decode (avg)
- 178.19 tok/s
- Prefill (avg)
- 583.74 tok/s
- TTFT (avg)
- 1.53 s
- VRAM (avg)
- 6.47 GB
- MTP acceptance rate
- —
- TTFB
- 1.43 GB
- Power draw
- 66.2 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B-Thinking-2507
- Quantization
- IQ2_M
- Framework
- llama.cpp
- Version
- v0.52
- Context length
- 8192 tokens
- Batch size
- 256
- 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
- AMD Ryzen AI Max+ 395
- Nominal VRAM
- 128 GB
- Measured VRAM (avg)
- 6.47 GB
- OS
- Windows 11 23H2
- Driver
- AMD Adrenalin 25.10.1
- CUDA
- N/A
- Power draw
- 66.2 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.52 · Windows 11 23H2 · CUDA N/A · 8192 ctx
178.19 tok/s
Decode
583.74 tok/s
Prefill
1.53 s
TTFT
6.47 GB
VRAM
—
MTP
66.2 W
Power draw
L2 batch seed 2026-10-13: llama.cpp on hw_id=24 running model_id=57 at IQ2_M