Calling all GPUs: help us benchmark SonicBoom v0.0.3

Thank you for helping! This issue collects results for v0.0.3 only. It will be locked when the next version's benchmark issue opens; results posted here go into the README community table with your GitHub handle.

1. Get the models

• Gemma 4 E4B: <HF repo>/gemma-4-E4B-it-Q3_K_M.gguf
• Qwen2.5 0.5B: https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF → qwen2.5-0.5b-instruct-q4_k_m.gguf

2. Build SonicBoom v0.0.3

git clone https://github.com/hardenedlinux/sonicboom && cd sonicboom
git checkout v0.0.3
<build commands exactly as in README>

3. (Optional) llama.cpp for the comparison column

Same commit we use: 43fe9c64281, built with CUDA; point LLAMA_BENCH at llama-bench.

4. Run

LLAMA_BENCH=/path/to/llama-bench tools/sb-report.sh \
  --gemma /path/to/gemma-4-E4B-it-Q3_K_M.gguf \
  --qwen  /path/to/qwen2.5-0.5b-instruct-q4_k_m.gguf

5. Paste

The whole Markdown block it prints (build info, GPU, golden checks, benchmark tables). If something fails, paste the output anyway — failures on hardware we don't have are the most valuable reports.