diff --git a/config/model-scores.yaml b/config/model-scores.yaml
index 032041e..a9a2222 100644
--- a/config/model-scores.yaml
+++ b/config/model-scores.yaml
@@ -77,6 +77,12 @@ model_scores:
     overall: 66.0
     reasoning: 72.0
     vision: 62.0
+  longcat-2.5-preview-free:
+    coding: 66.0             # no benchmarks published (released 2026-09-25); estimate: vendor claims improved coding over longcat-2.0, free tier reduction
+    instruction_following: 86.0  # estimate: longcat-2.0 IF 90, free tier reduction
+    overall: 55.0             # estimate: no LiveBench/AA data; conservative for preview free tier
+    reasoning: 84.0           # estimate: longcat-2.0 reasoning 89, free tier reduction
+    vision: 30.0              # vendor changelog claims image input; no vision benchmark data; conservative
   laguna-s-2.1-free:
     coding: 59.0              # SWE-bench Pro 59.4 (poolside/HF/BenchLM agree); T-Bench 70.2, SWE-Multi 78.5
     instruction_following: 54.0