add 4 20b+ models after enabling dynamic gpu duration

app.py

@@ -27,38 +27,22 @@ cancel_event = threading.Event()
 # ------------------------------
 MODELS = {
     # ~30.5B total parameters (MoE: 3.3B activated)
-    # "Qwen3-30B-A3B-Instruct-2507-AWQ-4bit": {
-    #     "repo_id": "cpatonn/Qwen3-30B-A3B-Instruct-2507-AWQ-4bit",
-    #     "description": "4-bit AWQ quantized instruct-tuned MoE model based on Qwen3-30B-A3B-Instruct-2507. Features 30.5B total parameters (3.3B activated), 128 experts (8 activated), 48 layers, and native 262,144-token context. Excels in instruction following, logical reasoning, multilingualism, coding, and long-context understanding. Supports only non-thinking mode (no <think> blocks). Quantized using AWQ (W4A16) with lm_head and gating layers preserved in higher precision."
-    # },
-    # "Qwen3-30B-A3B-Thinking-2507-AWQ-4bit": {
-    #     "repo_id": "cpatonn/Qwen3-30B-A3B-Thinking-2507-AWQ-4bit",
-    #     "description": "4-bit AWQ quantized thinking-mode MoE model based on Qwen3-30B-A3B-Thinking-2507. Contains 30.5B total parameters (3.3B activated), 128 experts (8 activated), 48 layers, and 262,144-token native context. Optimized for deep reasoning in mathematics, science, coding, and agent tasks. Outputs include automatic reasoning delimiters (<think>...</think>). Quantized with AWQ (W4A16), preserving lm_head and expert gating layers."
-    # },
-
-    # # ~80B total parameters (MoE: 3B activated)
-    # "Qwen3-Next-80B-A3B-Instruct-AWQ-4bit": {
-    #     "repo_id": "cpatonn/Qwen3-Next-80B-A3B-Instruct-AWQ-4bit",
-    #     "description": "4-bit AWQ quantized instruct model from the Qwen3-Next series. Features 80B total parameters (3B activated), hybrid Gated DeltaNet + Gated Attention architecture, 512 experts (10 activated + 1 shared), and native 262,144-token context (extendable to 1M tokens with YaRN). Delivers performance comparable to Qwen3-235B on many benchmarks while offering superior ultra-long-context efficiency. Supports only non-thinking mode. Note: May require re-quantization for stable inference (as of Sept 2025)."
-    # },
-    # "Qwen3-Next-80B-A3B-Thinking-AWQ-4bit": {
-    #     "repo_id": "cpatonn/Qwen3-Next-80B-A3B-Thinking-AWQ-4bit",
-    #     "description": "4-bit AWQ quantized thinking-mode variant of Qwen3-Next-80B-A3B. Combines 80B total parameters (3B activated), hybrid attention (Gated DeltaNet + Gated Attention), and 512-expert MoE (10 activated + 1 shared) for advanced reasoning over ultra-long contexts (natively 262K, extendable to 1M tokens). Designed for complex problem-solving with automatic reasoning trace generation. Quantized using AWQ; intended for high-end agentic and analytical workloads."
-    # },
-
-    # ~235B total parameters (MoE: 22B activated) — included for reference if added later
-    # "Qwen3-235B-A22B-Thinking": { ... },
-
-    # 32.8B total parameters
-    # "Qwen3-32B-AWQ": {
-    #     "repo_id": "Qwen/Qwen3-32B-AWQ",
-    #     "description": "4-bit AWQ quantized dense causal language model with 32.8B total parameters (31.2B non-embedding), 64 layers, 64 query heads & 8 KV heads, native 32,768-token context (extendable to 131,072 via YaRN). Features seamless switching between thinking mode (for complex reasoning, math, coding) and non-thinking mode (for efficient dialogue), strong multilingual support (100+ languages), and leading open-source agent capabilities."
-    # },
-
-    # "gpt-oss-20b-BF16": {
-    #     "repo_id": "unsloth/gpt-oss-20b-BF16",
-    #     "description": "A 20B-parameter open-source GPT-style language model quantized to INT4 using AutoRound, with FP8 key-value cache for efficient inference. Optimized for performance and memory efficiency on Intel hardware while maintaining strong language generation capabilities."
-    # },
+    "Qwen3-30B-A3B-Instruct-2507": {
+        "repo_id": "Qwen/Qwen3-30B-A3B-Instruct-2507",
+        "description": "non-thinking-mode MoE model based on Qwen3-30B-A3B-Instruct-2507. Features 30.5B total parameters (3.3B activated), 128 experts (8 activated), 48 layers, and native 262,144-token context. Excels in instruction following, logical reasoning, multilingualism, coding, and long-context understanding. Supports only non-thinking mode (no <think> blocks). Quantized using AWQ (W4A16) with lm_head and gating layers preserved in higher precision."
+    },
+    "Qwen3-30B-A3B-Thinking-2507": {
+        "repo_id": "Qwen/Qwen3-30B-A3B-Thinking-2507",
+        "description": "thinking-mode MoE model based on Qwen3-30B-A3B-Thinking-2507. Contains 30.5B total parameters (3.3B activated), 128 experts (8 activated), 48 layers, and 262,144-token native context. Optimized for deep reasoning in mathematics, science, coding, and agent tasks. Outputs include automatic reasoning delimiters (<think>...</think>). Quantized with AWQ (W4A16), preserving lm_head and expert gating layers."
+    },
+    "Qwen3-32B-FP8": {
+        "repo_id": "Qwen/Qwen3-32B-FP8",
+        "description": "Dense causal language model with 32.8B total parameters (31.2B non-embedding), 64 layers, 64 query heads & 8 KV heads, native 32,768-token context (extendable to 131,072 via YaRN). Features seamless switching between thinking mode (for complex reasoning, math, coding) and non-thinking mode (for efficient dialogue), strong multilingual support (100+ languages), and leading open-source agent capabilities."
+    },
+    "gpt-oss-20b-BF16": {
+        "repo_id": "unsloth/gpt-oss-20b-BF16",
+        "description": "A 20B-parameter open-source GPT-style language model quantized to INT4 using AutoRound, with FP8 key-value cache for efficient inference. Optimized for performance and memory efficiency on Intel hardware while maintaining strong language generation capabilities."
+    },
     "Qwen3-4B-Instruct-2507": {
         "repo_id": "Qwen/Qwen3-4B-Instruct-2507",
         "description": "Updated non-thinking instruct variant of Qwen3-4B with 4.0B parameters, featuring significant improvements in instruction following, logical reasoning, multilingualism, and 256K long-context understanding. Strong performance across knowledge, coding, alignment, and agent benchmarks."