Upload 2 files

README.md

@@ -11,7 +11,7 @@ hf_oauth: true
 hf_oauth_scopes:
 - inference-api
 license: apache-2.0
-short_description: Compact model arena with Qwen commentary
+short_description: Compact model arena with GLM and GPT OSS commentary
 models:
 - HuggingFaceTB/SmolLM2-135M
 - MaliosDark/Isabel-50M
@@ -51,7 +51,6 @@ models:
 - joelhenwang/OdinNext-138M-Instruct
 - veyra-ai/Veyra2-30M-Instruct-Early
 - MinimaLabs/KeyLM-75M-Instruct
-- Qwen/Qwen3.6-35B-A3B-FP8
 tags:
 - text-generation
 - small-language-model

app.py

@@ -1395,8 +1395,7 @@ with gr.Blocks(title=APP_TITLE, css=CSS, fill_width=True) as demo:
           <p>
             Compare two to five compact Hugging Face models side by side. You can pick visible models,
             run a blind manual matchup, or ask the arena to randomize hidden contenders with organization
-            and size-band rules. AI commentary runs locally on CPU with a preloaded Qwen/Qwen3.5-2B
-            commentator.
+            and size-band rules. AI commentary runs through GLM 4.7 or GPT OSS 120B on CPU.
           </p>
           <div class="micro-grid">
             <div class="micro-card"><strong>Mode 1</strong><span>Pick and see the models while they answer.</span></div>