Add claude 4.6 note

app.py

@@ -103,6 +103,12 @@ ArmBench-LLM is a comprehensive benchmark for evaluating large language models o
             markdown += f"  - {display_name} ({metric_desc})\n"
 
     markdown += """
+## Important Note on Claude 4.6 Models
+
+We observed significant reliability issues with Claude 4.6 during evaluation, which produced inconsistent and non-reproducible results. To preserve the integrity of the benchmark, all Claude 4.6 models have been excluded from the leaderboard.
+
+You can learn more about the issue in [this GitHub issue](https://github.com/anthropics/claude-code/issues/17357).
+
 ## Submission Guide
 
 To submit your model for evaluation:

model_sizes.csv

@@ -23,3 +23,5 @@ openai/gpt-5.4-mini, -
 openai/gpt-5.4-nano, - 
 x-ai/grok-4.20-beta, - 
 qwen/qwen3.5-27b,27B
+anthropic/claude-3.7-sonnet, - 
+anthropic/claude-sonnet-4, -