Trying Qwen/Qwen2.5-14B-Instruct model to test performance

config.py

@@ -1,8 +1,7 @@
-# ── LLM Backend ──────────────────────────────────────────────────────────────
-# ── LLM Backend ──────────────────────────────────────────────────────────────
-#─────────────────────────────────────────────────────────
 # Set to a globally maintained, high-throughput serverless instruction model
-HF_MODEL = "Qwen/Qwen2.5-7B-Instruct"
+# ── LLM Backend ──────────────────────────────────────────────────────────────
+# Set to the high-availability 14B Qwen cluster for context-aware grading
+HF_MODEL = "Qwen/Qwen2.5-14B-Instruct"
 
 # ── Persistence ──────────────────────────────────────────────────────────────
 # Session is runtime-only. History JSON is written per-session and read for PDF.