diff --git a/scripts/vllm-awq-qwen36-dflash-ampere.sh b/scripts/vllm-awq-qwen36-dflash-ampere.sh
index dc464af..23e86a3 100644
--- a/scripts/vllm-awq-qwen36-dflash-ampere.sh
+++ b/scripts/vllm-awq-qwen36-dflash-ampere.sh
@@ -23,7 +23,6 @@ exec docker run --rm --name vllm-awq-qwen36-dflash-ampere \
     -v ~/.cache/huggingface:/root/.cache/huggingface \
     vllm/vllm-openai:nightly "${MODEL}" \
     --served-model-name "frank-default-small" \
-    # --speculative-config '{"method": "dflash", "model": "z-lab/Qwen3.6-35B-A3B-DFlash", "num_speculative_tokens": 12}' \
     --max-num-seqs 2 \
     --max-model-len 32768 \
     --tensor-parallel-size 1 \
@@ -33,7 +32,10 @@ exec docker run --rm --name vllm-awq-qwen36-dflash-ampere \
     --enable-auto-tool-choice --tool-call-parser qwen3_xml \
     --reasoning-parser qwen3 \
     --trust-remote-code \
-    --gpu-memory-utilization 0.75 \
+    --gpu-memory-utilization 0.95 \
     --cpu-offload-gb 20 \
     --cpu-offload-params "experts" \
     --default-chat-template-kwargs '{"enable_thinking": true, "preserve_thinking": true}'
+
+# TODO: uncomment and move to after --served-model-name to enable DFlash
+#    --speculative-config '{"method": "dflash", "model": "z-lab/Qwen3.6-35B-A3B-DFlash", "num_speculative_tokens": 12}' \