app.py

import gradio as gr
import gemini_gradio
import openai_gradio
import anthropic_gradio
import sambanova_gradio
import xai_gradio
import hyperbolic_gradio
import perplexity_gradio
import mistral_gradio
import fireworks_gradio
import cerebras_gradio
import groq_gradio
import together_gradio
import nvidia_gradio
import dashscope_gradio



with gr.Blocks(fill_height=True) as demo:
    with gr.Tab("Meta Llama"):
        with gr.Row():
            llama_model = gr.Dropdown(
                choices=[
                    'Meta-Llama-3.2-1B-Instruct',   # Llama 3.2 1B
                    'Meta-Llama-3.2-3B-Instruct',   # Llama 3.2 3B
                    'Llama-3.2-11B-Vision-Instruct',  # Llama 3.2 11B
                    'Llama-3.2-90B-Vision-Instruct',  # Llama 3.2 90B
                    'Meta-Llama-3.1-8B-Instruct',    # Llama 3.1 8B
                    'Meta-Llama-3.1-70B-Instruct',   # Llama 3.1 70B
                    'Meta-Llama-3.1-405B-Instruct'   # Llama 3.1 405B
                ],
                value='Llama-3.2-90B-Vision-Instruct',  # Default to the most advanced model
                label="Select Llama Model",
                interactive=True
            )
        
        llama_interface = gr.load(
            name=llama_model.value,
            src=sambanova_gradio.registry,
            multimodal=True,
            fill_height=True
        )
        
        def update_llama_model(new_model):
            return gr.load(
                name=new_model,
                src=sambanova_gradio.registry,
                multimodal=True,
                fill_height=True
            )
        
        llama_model.change(
            fn=update_llama_model,
            inputs=[llama_model],
            outputs=[llama_interface]
        )
        
        gr.Markdown("**Note:** You need to use a SambaNova API key from [SambaNova Cloud](https://cloud.sambanova.ai/).")
    with gr.Tab("Gemini"):
        with gr.Row():
            gemini_model = gr.Dropdown(
                choices=[
                    'gemini-1.5-flash',        # Fast and versatile performance
                    'gemini-1.5-flash-8b',     # High volume, lower intelligence tasks
                    'gemini-1.5-pro',           # Complex reasoning tasks
                    'gemini-exp-1114'          # Quality improvements
                ],
                value='gemini-1.5-pro',      # Default to the most advanced model
                label="Select Gemini Model",
                interactive=True
            )
        
        gemini_interface = gr.load(
            name=gemini_model.value,
            src=gemini_gradio.registry,
            fill_height=True
        )
        
        def update_gemini_model(new_model):
            return gr.load(
                name=new_model,
                src=gemini_gradio.registry,
                fill_height=True
            )
        
        gemini_model.change(
            fn=update_gemini_model,
            inputs=[gemini_model],
            outputs=[gemini_interface]
        )
    with gr.Tab("ChatGPT"):
        with gr.Row():
            model_choice = gr.Dropdown(
                choices=[
                    'gpt-4o-2024-11-20',              # Latest GPT-4o model
                    'gpt-4o',                         # Previous most advanced model
                    'gpt-4o-2024-08-06',              # Latest snapshot
                    'gpt-4o-2024-05-13',              # Original snapshot
                    'chatgpt-4o-latest',          # Latest ChatGPT version
                    'gpt-4o-mini',                # Small model
                    'gpt-4o-mini-2024-07-18',     # Latest mini version
                    'o1-preview',                 # Reasoning model
                    'o1-preview-2024-09-12',      # Latest o1 model snapshot
                    'o1-mini',                    # Faster reasoning model
                    'o1-mini-2024-09-12',         # Latest o1-mini model snapshot
                    'gpt-4-turbo',                # Latest GPT-4 Turbo model
                    'gpt-4-turbo-2024-04-09',     # Latest GPT-4 Turbo snapshot
                    'gpt-4-turbo-preview',         # GPT-4 Turbo preview model
                    'gpt-4-0125-preview',         # GPT-4 Turbo preview model for laziness
                    'gpt-4-1106-preview',         # Improved instruction following model
                    'gpt-4',                      # Standard GPT-4 model
                    'gpt-4-0613'                  # Snapshot of GPT-4 from June 2023
                ],
                value='gpt-4o-2024-11-20',           # Updated default to latest model
                label="Select Model",
                interactive=True
            )
            
        chatgpt_interface = gr.load(
            name=model_choice.value,
            src=openai_gradio.registry,
            fill_height=True
        )
        
        def update_model(new_model):
            return gr.load(
                name=new_model,
                src=openai_gradio.registry,
                fill_height=True
            )
        
        model_choice.change(
            fn=update_model,
            inputs=[model_choice],
            outputs=[chatgpt_interface]
        )
    with gr.Tab("Claude"):
        with gr.Row():
            claude_model = gr.Dropdown(
                choices=[
                    'claude-3-5-sonnet-20241022',  # Latest Sonnet
                    'claude-3-5-haiku-20241022',   # Latest Haiku
                    'claude-3-opus-20240229',       # Opus
                    'claude-3-sonnet-20240229',     # Previous Sonnet
                    'claude-3-haiku-20240307'       # Previous Haiku
                ],
                value='claude-3-5-sonnet-20241022',  # Default to latest Sonnet
                label="Select Model",
                interactive=True
            )
            
        claude_interface = gr.load(
            name=claude_model.value,
            src=anthropic_gradio.registry,
            accept_token=True,
            fill_height=True
        )
        
        def update_claude_model(new_model):
            return gr.load(
                name=new_model,
                src=anthropic_gradio.registry,
                accept_token=True,
                fill_height=True
            )
        
        claude_model.change(
            fn=update_claude_model,
            inputs=[claude_model],
            outputs=[claude_interface]
        )
    with gr.Tab("Grok"):
        with gr.Row():
            grok_model = gr.Dropdown(
                choices=[
                    'grok-beta',
                    'grok-vision-beta'
                ],
                value='grok-vision-beta',
                label="Select Grok Model",
                interactive=True
            )
            
        grok_interface = gr.load(
            name=grok_model.value,
            src=xai_gradio.registry,
            fill_height=True
        )
        
        def update_grok_model(new_model):
            return gr.load(
                name=new_model,
                src=xai_gradio.registry,
                fill_height=True
            )
        
        grok_model.change(
            fn=update_grok_model,
            inputs=[grok_model],
            outputs=[grok_interface]
        )
    with gr.Tab("Hugging Face"):
        with gr.Row():
            hf_model = gr.Dropdown(
                choices=[
                    # Latest Large Models
                    'Qwen/Qwen2.5-Coder-32B-Instruct',
                    'Qwen/Qwen2.5-72B-Instruct',
                    'meta-llama/Llama-3.1-70B-Instruct',
                    'mistralai/Mixtral-8x7B-Instruct-v0.1',
                    # Mid-size Models
                    'meta-llama/Llama-3.1-8B-Instruct',
                    'google/gemma-2-9b-it',
                    'mistralai/Mistral-7B-v0.1',
                    'meta-llama/Llama-2-7b-chat-hf',
                    # Smaller Models
                    'meta-llama/Llama-3.2-3B-Instruct',
                    'meta-llama/Llama-3.2-1B-Instruct',
                    'Qwen/Qwen2.5-1.5B-Instruct',
                    'microsoft/Phi-3.5-mini-instruct',
                    'HuggingFaceTB/SmolLM2-1.7B-Instruct',
                    'google/gemma-2-2b-it',
                    # Base Models
                    'meta-llama/Llama-3.2-3B',
                    'meta-llama/Llama-3.2-1B',
                    'openai-community/gpt2'
                ],
                value='HuggingFaceTB/SmolLM2-1.7B-Instruct',  # Default to a powerful model
                label="Select Hugging Face Model",
                interactive=True
            )
            
        hf_interface = gr.load(
            name=hf_model.value,
            src="models",  # Use direct model loading from HF
            fill_height=True
        )
        
        def update_hf_model(new_model):
            return gr.load(
                name=new_model,
                src="models",
                fill_height=True
            )
        
        hf_model.change(
            fn=update_hf_model,
            inputs=[hf_model],
            outputs=[hf_interface]
        )
        
        gr.Markdown("""
        **Note:** These models are loaded directly from Hugging Face Hub. Some models may require authentication.
        
        Models are organized by size:
        - **Large Models**: 32B-72B parameters
        - **Mid-size Models**: 7B-9B parameters
        - **Smaller Models**: 1B-3B parameters
        - **Base Models**: Original architectures
        
        Visit [Hugging Face](https://huggingface.co/) to learn more about available models.
        """)
    with gr.Tab("Groq"):
        with gr.Row():
            groq_model = gr.Dropdown(
                choices=[
                    'llama3-groq-8b-8192-tool-use-preview',
                    'llama3-groq-70b-8192-tool-use-preview',
                    'llama-3.2-1b-preview',
                    'llama-3.2-3b-preview',
                    'llama-3.2-11b-text-preview',
                    'llama-3.2-90b-text-preview',
                    'mixtral-8x7b-32768',
                    'gemma2-9b-it',
                    'gemma-7b-it'
                ],
                value='llama3-groq-70b-8192-tool-use-preview',  # Default to Groq's optimized model
                label="Select Groq Model",
                interactive=True
            )
            
        groq_interface = gr.load(
            name=groq_model.value,
            src=groq_gradio.registry,
            fill_height=True
        )
        
        def update_groq_model(new_model):
            return gr.load(
                name=new_model,
                src=groq_gradio.registry,
                fill_height=True
            )
        
        groq_model.change(
            fn=update_groq_model,
            inputs=[groq_model],
            outputs=[groq_interface]
        )
        
        gr.Markdown("""
        **Note:** You need a Groq API key to use these models. Get one at [Groq Cloud](https://console.groq.com/).
        """)
    with gr.Tab("Hyperbolic"):
        with gr.Row():
            hyperbolic_model = gr.Dropdown(
                choices=[
                    # # Vision Models (TODO)
                    # 'Qwen/Qwen2-VL-72B-Instruct',                       # 32K context
                    # 'mistralai/Pixtral-12B-2409',                       # 32K context
                    # 'Qwen/Qwen2-VL-7B-Instruct',                        # 32K context
                    # Large Language Models
                    'Qwen/Qwen2.5-Coder-32B-Instruct',                  # 131K context
                    'meta-llama/Llama-3.2-3B-Instruct',                 # 131K context
                    'meta-llama/Meta-Llama-3.1-8B-Instruct',            # 131k context
                    'meta-llama/Meta-Llama-3.1-70B-Instruct',           # 32K context
                    'meta-llama/Meta-Llama-3-70B-Instruct',             # 8K context
                    'NousResearch/Hermes-3-Llama-3.1-70B',              # 12K context
                    'Qwen/Qwen2.5-72B-Instruct',                        # 32K context
                    'deepseek-ai/DeepSeek-V2.5',                        # 8K context
                    'meta-llama/Meta-Llama-3.1-405B-Instruct',          # 8K context
                ],
                value='Qwen/Qwen2.5-Coder-32B-Instruct',
                label="Select Hyperbolic Model",
                interactive=True
            )
            
        hyperbolic_interface = gr.load(
            name=hyperbolic_model.value,
            src=hyperbolic_gradio.registry,
            fill_height=True
        )
        
        def update_hyperbolic_model(new_model):
            return gr.load(
                name=new_model,
                src=hyperbolic_gradio.registry,
                fill_height=True
            )
        
        hyperbolic_model.change(
            fn=update_hyperbolic_model,
            inputs=[hyperbolic_model],
            outputs=[hyperbolic_interface]
        )
        
        gr.Markdown("""
        <div>
            <img src="https://storage.googleapis.com/public-arena-asset/hyperbolic_logo.png" alt="Hyperbolic Logo" style="height: 50px; margin-right: 10px;">
        </div>    
                    
        **Note:** This model is supported by Hyperbolic. Build your AI apps at [Hyperbolic](https://app.hyperbolic.xyz/).
        """)
    with gr.Tab("Qwen"):
        with gr.Row():
            qwen_model = gr.Dropdown(
                choices=[
                    # Proprietary Qwen Models
                    'qwen-turbo-latest',
                    'qwen-turbo',
                    'qwen-plus',
                    'qwen-max',
                    # Open Source Qwen Models
                    'qwen1.5-110b-chat',
                    'qwen1.5-72b-chat',
                    'qwen1.5-32b-chat',
                    'qwen1.5-14b-chat',
                    'qwen1.5-7b-chat'
                ],
                value='qwen-turbo-latest',  # Default to the latest turbo model
                label="Select Qwen Model",
                interactive=True
            )
            
        qwen_interface = gr.load(
            name=qwen_model.value,
            src=dashscope_gradio.registry,
            fill_height=True
        )
        
        def update_qwen_model(new_model):
            return gr.load(
                name=new_model,
                src=dashscope_gradio.registry,
                fill_height=True
            )
        
        qwen_model.change(
            fn=update_qwen_model,
            inputs=[qwen_model],
            outputs=[qwen_interface]
        )
        
        gr.Markdown("""
        **Note:** You need a DashScope API key to use these models. Get one at [DashScope](https://dashscope.aliyun.com/).
        
        Models available in two categories:
        - **Proprietary Models**:
          - Qwen Turbo: Fast responses for general tasks
          - Qwen Plus: Balanced performance and quality
          - Qwen Max: Highest quality responses
        - **Open Source Models**:
          - Available in various sizes from 7B to 110B parameters
          - Based on the Qwen 1.5 architecture
        """)
    with gr.Tab("Perplexity"):
        with gr.Row():
            perplexity_model = gr.Dropdown(
                choices=[
                    # Sonar Models (Online)
                    'llama-3.1-sonar-small-128k-online',    # 8B params
                    'llama-3.1-sonar-large-128k-online',    # 70B params
                    'llama-3.1-sonar-huge-128k-online',     # 405B params
                    # Sonar Models (Chat)
                    'llama-3.1-sonar-small-128k-chat',      # 8B params
                    'llama-3.1-sonar-large-128k-chat',      # 70B params
                    # Open Source Models
                    'llama-3.1-8b-instruct',                # 8B params
                    'llama-3.1-70b-instruct'                # 70B params
                ],
                value='llama-3.1-sonar-large-128k-online',  # Default to large online model
                label="Select Perplexity Model",
                interactive=True
            )
        
        perplexity_interface = gr.load(
            name=perplexity_model.value,
            src=perplexity_gradio.registry,
            accept_token=True,
            fill_height=True
        )
        
        def update_perplexity_model(new_model):
            return gr.load(
                name=new_model,
                src=perplexity_gradio.registry,
                accept_token=True,
                fill_height=True
            )
        
        perplexity_model.change(
            fn=update_perplexity_model,
            inputs=[perplexity_model],
            outputs=[perplexity_interface]
        )
        
        gr.Markdown("""
        **Note:** Models are grouped into three categories:
        - **Sonar Online Models**: Include search capabilities (beta access required)
        - **Sonar Chat Models**: Standard chat models
        - **Open Source Models**: Based on Hugging Face implementations
        
        For access to Online LLMs features, please fill out the [beta access form](https://perplexity.typeform.com/apiaccessform?typeform-source=docs.perplexity.ai).
        """)
    with gr.Tab("DeepSeek-V2.5"):
        gr.load(
            name='deepseek-ai/DeepSeek-V2.5',
            src=hyperbolic_gradio.registry,
            fill_height=True
        )
        gr.Markdown("""
        <div>
            <img src="https://storage.googleapis.com/public-arena-asset/hyperbolic_logo.png" alt="Hyperbolic Logo" style="height: 50px; margin-right: 10px;">
        </div>    
                    
        **Note:** This model is supported by Hyperbolic. Build your AI apps at [Hyperbolic](https://app.hyperbolic.xyz/).
        """)
    with gr.Tab("Mistral"):
        with gr.Row():
            mistral_model = gr.Dropdown(
                choices=[
                    # Premier Models
                    'mistral-large-latest',          # Top-tier reasoning model (128k)
                    'pixtral-large-latest',          # Frontier-class multimodal model (128k)
                    'ministral-3b-latest',           # Best edge model (128k)
                    'ministral-8b-latest',           # High performance edge model (128k)
                    'mistral-small-latest',          # Enterprise-grade small model (32k)
                    'codestral-latest',              # Code-specialized model (32k)
                    'mistral-embed',                 # Semantic text representation (8k)
                    'mistral-moderation-latest',     # Content moderation service (8k)
                    # Free Models
                    'pixtral-12b-2409',             # Free 12B multimodal model (128k)
                    'open-mistral-nemo',             # Multilingual model (128k)
                    'open-codestral-mamba'           # Mamba-based coding model (256k)
                ],
                value='pixtral-large-latest',    # pixtral for vision
                label="Select Mistral Model",
                interactive=True
            )
            
        mistral_interface = gr.load(
            name=mistral_model.value,
            src=mistral_gradio.registry,
            fill_height=True
        )
        
        def update_mistral_model(new_model):
            return gr.load(
                name=new_model,
                src=mistral_gradio.registry,
                fill_height=True
            )
        
        mistral_model.change(
            fn=update_mistral_model,
            inputs=[mistral_model],
            outputs=[mistral_interface],
        )
        
        gr.Markdown("""
        **Note:** You need a Mistral API key to use these models. Get one at [Mistral AI Platform](https://console.mistral.ai/).
        
        Models are grouped into two categories:
        - **Premier Models**: Require a paid API key
        - **Free Models**: Available with free API keys
        
        Each model has different context window sizes (from 8k to 256k tokens) and specialized capabilities.
        """)
    with gr.Tab("Fireworks"):
        with gr.Row():
            fireworks_model = gr.Dropdown(
                choices=[
                    'f1-preview',              # Latest F1 preview model
                    'f1-mini-preview',         # Smaller, faster model
                ],
                value='f1-preview',            # Default to preview model
                label="Select Fireworks Model",
                interactive=True
            )
            
        fireworks_interface = gr.load(
            name=fireworks_model.value,
            src=fireworks_gradio.registry,
            fill_height=True
        )
        
        def update_fireworks_model(new_model):
            return gr.load(
                name=new_model,
                src=fireworks_gradio.registry,
                fill_height=True
            )
        
        fireworks_model.change(
            fn=update_fireworks_model,
            inputs=[fireworks_model],
            outputs=[fireworks_interface]
        )
        
        gr.Markdown("""
        **Note:** You need a Fireworks AI API key to use these models. Get one at [Fireworks AI](https://app.fireworks.ai/).
        """)
    with gr.Tab("Cerebras"):
        with gr.Row():
            cerebras_model = gr.Dropdown(
                choices=[
                    'llama3.1-8b',
                    'llama3.1-70b',
                    'llama3.1-405b'
                ],
                value='llama3.1-70b',  # Default to mid-size model
                label="Select Cerebras Model",
                interactive=True
            )
            
        cerebras_interface = gr.load(
            name=cerebras_model.value,
            src=cerebras_gradio.registry,
            accept_token=True,  # Added token acceptance
            fill_height=True
        )
        
        def update_cerebras_model(new_model):
            return gr.load(
                name=new_model,
                src=cerebras_gradio.registry,
                accept_token=True,  # Added token acceptance
                fill_height=True
            )
        
        cerebras_model.change(
            fn=update_cerebras_model,
            inputs=[cerebras_model],
            outputs=[cerebras_interface]
        )
    with gr.Tab("Together"):
        with gr.Row():
            together_model = gr.Dropdown(
                choices=[
                    # Vision Models
                    'meta-llama/Llama-Vision-Free',                     # 131k context (Free)
                    'meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo',  # 131k context
                    'meta-llama/Llama-3.2-90B-Vision-Instruct-Turbo',  # 131k context
                    # Meta Llama 3.x Models
                    'meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo',      # 131k context
                    'meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo',     # 131k context
                    'meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo',    # 130k context
                    'meta-llama/Meta-Llama-3-8B-Instruct-Turbo',        # 8k context
                    'meta-llama/Meta-Llama-3-70B-Instruct-Turbo',       # 8k context
                    'meta-llama/Llama-3.2-3B-Instruct-Turbo',          # 131k context
                    'meta-llama/Meta-Llama-3-8B-Instruct-Lite',         # 8k context, INT4
                    'meta-llama/Meta-Llama-3-70B-Instruct-Lite',        # 8k context, INT4
                    'meta-llama/Llama-3-8b-chat-hf',                    # 8k context
                    'meta-llama/Llama-3-70b-chat-hf',                   # 8k context
                    # Other Large Models
                    'nvidia/Llama-3.1-Nemotron-70B-Instruct-HF',        # 32k context
                    'Qwen/Qwen2.5-Coder-32B-Instruct',                  # 32k context
                    'microsoft/WizardLM-2-8x22B',                       # 65k context
                    'google/gemma-2-27b-it',                            # 8k context
                    'google/gemma-2-9b-it',                             # 8k context
                    'databricks/dbrx-instruct',                         # 32k context
                    # Mixtral Models
                    'mistralai/Mixtral-8x7B-Instruct-v0.1',            # 32k context
                    'mistralai/Mixtral-8x22B-Instruct-v0.1',           # 65k context
                    # Qwen Models
                    'Qwen/Qwen2.5-7B-Instruct-Turbo',                  # 32k context
                    'Qwen/Qwen2.5-72B-Instruct-Turbo',                 # 32k context
                    'Qwen/Qwen2-72B-Instruct',                         # 32k context
                    # Other Models
                    'deepseek-ai/deepseek-llm-67b-chat',               # 4k context
                    'google/gemma-2b-it',                              # 8k context
                    'Gryphe/MythoMax-L2-13b',                          # 4k context
                    'meta-llama/Llama-2-13b-chat-hf',                  # 4k context
                    'mistralai/Mistral-7B-Instruct-v0.1',              # 8k context
                    'mistralai/Mistral-7B-Instruct-v0.2',              # 32k context
                    'mistralai/Mistral-7B-Instruct-v0.3',              # 32k context
                    'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO',     # 32k context
                    'togethercomputer/StripedHyena-Nous-7B',           # 32k context
                    'upstage/SOLAR-10.7B-Instruct-v1.0'                # 4k context
                ],
                value='meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo',  # Default to recommended vision model
                label="Select Together Model",
                interactive=True
            )
            
        together_interface = gr.load(
            name=together_model.value,
            src=together_gradio.registry,
            multimodal=True,
            fill_height=True
        )
        
        def update_together_model(new_model):
            return gr.load(
                name=new_model,
                src=together_gradio.registry,
                multimodal=True,
                fill_height=True
            )
        
        together_model.change(
            fn=update_together_model,
            inputs=[together_model],
            outputs=[together_interface]
        )
        
        gr.Markdown("""
        **Note:** You need a Together AI API key to use these models. Get one at [Together AI](https://www.together.ai/).
        """)
    with gr.Tab("NVIDIA"):
        with gr.Row():
            nvidia_model = gr.Dropdown(
                choices=[
                    # NVIDIA Models
                    'nvidia/llama3-chatqa-1.5-70b',
                    'nvidia/llama3-chatqa-1.5-8b',
                    'nvidia-nemotron-4-340b-instruct',
                    # Meta Models
                    'meta/llama-3.1-70b-instruct',    # Added Llama 3.1 70B
                    'meta/codellama-70b',
                    'meta/llama2-70b',
                    'meta/llama3-8b',
                    'meta/llama3-70b',
                    # Mistral Models
                    'mistralai/codestral-22b-instruct-v0.1',
                    'mistralai/mathstral-7b-v0.1',
                    'mistralai/mistral-large-2-instruct',
                    'mistralai/mistral-7b-instruct',
                    'mistralai/mistral-7b-instruct-v0.3',
                    'mistralai/mixtral-8x7b-instruct',
                    'mistralai/mixtral-8x22b-instruct',
                    'mistralai/mistral-large',
                    # Google Models
                    'google/gemma-2b',
                    'google/gemma-7b',
                    'google/gemma-2-2b-it',
                    'google/gemma-2-9b-it',
                    'google/gemma-2-27b-it',
                    'google/codegemma-1.1-7b',
                    'google/codegemma-7b',
                    'google/recurrentgemma-2b',
                    'google/shieldgemma-9b',
                    # Microsoft Phi-3 Models
                    'microsoft/phi-3-medium-128k-instruct',
                    'microsoft/phi-3-medium-4k-instruct',
                    'microsoft/phi-3-mini-128k-instruct',
                    'microsoft/phi-3-mini-4k-instruct',
                    'microsoft/phi-3-small-128k-instruct',
                    'microsoft/phi-3-small-8k-instruct',
                    # Other Models
                    'qwen/qwen2-7b-instruct',
                    'databricks/dbrx-instruct',
                    'deepseek-ai/deepseek-coder-6.7b-instruct',
                    'upstage/solar-10.7b-instruct',
                    'snowflake/arctic'
                ],
                value='meta/llama-3.1-70b-instruct',  # Changed default to Llama 3.1 70B
                label="Select NVIDIA Model",
                interactive=True
            )
            
        nvidia_interface = gr.load(
            name=nvidia_model.value,
            src=nvidia_gradio.registry,
            accept_token=True,
            fill_height=True
        )
        
        def update_nvidia_model(new_model):
            return gr.load(
                name=new_model,
                src=nvidia_gradio.registry,
                accept_token=True,
                fill_height=True
            )
        
        nvidia_model.change(
            fn=update_nvidia_model,
            inputs=[nvidia_model],
            outputs=[nvidia_interface]
        )
        
        gr.Markdown("""
        **Note:** You need an NVIDIA AI Foundation API key to use these models. Get one at [NVIDIA AI Foundation](https://www.nvidia.com/en-us/ai-data-science/foundation-models/).
        
        Models are organized by provider:
        - **NVIDIA**: Native models including Llama3-ChatQA and Nemotron
        - **Meta**: Llama family models
        - **Mistral**: Various Mistral and Mixtral models
        - **Google**: Gemma family models
        - **Microsoft**: Phi-3 series
        - And other providers including Qwen, Databricks, DeepSeek, etc.
        """)

demo.launch(ssr_mode=False)