feat: Update R2-Router to 6-model Global KNN (K=80, λ=0.999)

- Models: 235b, 80b, 30b, coder-next, gemini-flash, haiku
- Budgets: concise, budget_200, budget_400, budget_800
- Training: sub_10 only (809 queries), Global KNN (cosine, distance-weighted)
- Results: Acc=71.20%, Cost=$0.037/1kq, Arena(β=0.1)=71.94
- Beats Method 1 and Method 2 across all β values

Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>

Author: 3 people

router_inference/config/r2-router.json

@@ -4,10 +4,13 @@
     "router_cls_name": "R2Router",
     "models": [
       "qwen/qwen3-235b-a22b-2507",
+      "qwen/qwen3-next-80b-a3b-instruct",
+      "qwen/qwen3-30b-a3b-instruct-2507",
+      "Qwen/Qwen3-Coder-Next",
       "gemini-2.5-flash",
-      "mistralai/ministral-3-3b-2512"
+      "claude-3-haiku-20240307"
     ],
-    "description": "R2-Router: Global KNN (K=28, cosine, distance-weighted) trained on sub_10 split. Routes via risk=(1-lambda)*quality - lambda*tokens*price. Lambda=0.85.",
-    "lambda": 0.85
+    "description": "R2-Router: Global KNN (K=80, cosine, distance-weighted) trained on sub_10 split. 6 models, 4 budgets (concise, budget_200, budget_400, budget_800). Routes via risk=(1-lambda)*quality - lambda*tokens*price. Lambda=0.999.",
+    "lambda": 0.999
   }
 }