Qwen3.8-27B with MTP on RTX 4090: runs are posted

Runs we already have

• r_9yhg3wvngab · RTX 4090 (24GB) · 75.3 tok/s decode (chat-short) · measured by llm-speed (@llmspeed)
• r_7j8rpk9mwhg · RTX 4090 (24GB) · 84.5 tok/s decode (chat-short) · community run
• MTP-off control: r_do9h933nvme · RTX 4090 (24GB) · 43.1 tok/s decode (chat-short) · measured by llm-speed (@llmspeed) · same llama.cpp build and quant as r_9yhg3wvngab

Run it yourself

Start llama-server with the model (flags below are a starting point; adjust offload to your machine), then point llm-speed at it. Add --llama-args='...' with the flags you used so the run records them.

pipx install llm-speed   # or: pipx upgrade llm-speed
# any recent llama.cpp release
llama-server -m /path/to/MODEL.gguf --port 8080 -ngl 999
# then, in a second terminal:
LLAMA_CPP_SERVER_URL=http://127.0.0.1:8080 \
  llm-speed bench --backend llama.cpp \
  --model /path/to/MODEL.gguf \
  --workload chat-short

The upload is signed and public. How contributing works · what the CLI sends.