diff --git a/.env b/.env
index 1d1d8b9..b876891 100644
--- a/.env
+++ b/.env
@@ -4,8 +4,8 @@ MODEL_PATH=/models/deepseek-coder-33b-gptq
 SERVED_NAME=deepseek-coder
 TP_SIZE=4
 DTYPE=float16
-MAX_MODEL_LEN=8192
-GPU_MEM_UTIL=0.88
+MAX_MODEL_LEN=65536
+GPU_MEM_UTIL=0.60 # 0.88
 MAX_NUM_SEQS=4
 VLLM_EXTRA_ARGS=--quantization gptq --disable-custom-all-reduce --enforce-eager
 
diff --git a/compose.yaml b/compose.yaml
index 8fd33d7..48e87ae 100644
--- a/compose.yaml
+++ b/compose.yaml
@@ -8,7 +8,7 @@ services:
     env_file:
       - .env
     volumes:
-      - models:/models:ro
+      - /var/models:/models:ro
     deploy:
       resources:
         reservations:
@@ -22,7 +22,7 @@ services:
         exec python3 -m vllm.entrypoints.openai.api_server
         --model "$$MODEL_PATH"
         --served-model-name "$$SERVED_NAME"
-        --host 0.0.0.0 --port 8000
+        --host 0.0.0.0 --port "$LISTEN_PORT"
         --tensor-parallel-size "$$TP_SIZE"
         --dtype "$$DTYPE"
         --max-model-len "$$MAX_MODEL_LEN"