Models

Browse, fork, and train AI models on the HUD platform.

The Models page at hud.ai/models lets you browse available models, track your trained model checkpoints, and view inference logs. Models are the AI providers you use through the HUD Gateway—Claude, GPT, Gemini, and more—plus any custom models you’ve trained.

Overview

Navigate to Models to see two tabs:

• Explore — Browse public base models from providers (Anthropic, OpenAI, Google, xAI)
• My Models — Your forked or trained models

Model Details

Click on any model to see its detail page:

Checkpoints Tab

Shows the checkpoint tree for your model. Each checkpoint represents a saved state during training:

• HEAD — The active checkpoint used for inference
• Tree View — Visual history of training branches
• Click a checkpoint — View details, set as HEAD, or start new training

Traces Tab

View all traces where this model was used:

• Filter by checkpoint
• Click to view full trace details
• See prompts, tool calls, and responses

Logs Tab

Inference logs for API calls through the Gateway:

• Request/response details
• Token usage
• Latency metrics
• Filter by checkpoint

Settings Tab

Model configuration:

• Display Name — How the model appears in the UI
• Advanced — Model ID, API name, provider, routes, and other metadata

Training Models

To train a model, you need a base model to start from:

1. Go to Explore and find a trainable model
2. Click Fork to create your own copy
3. Click Train Model to start a training job

Training creates new checkpoints that appear in your checkpoint tree. You can set any checkpoint as HEAD to use it for inference.

Not all models are trainable. Look for the “Train Model” button to be enabled.

Forking Models

Fork a model to create your own copy that you can train:

1. Navigate to the model you want to fork
2. Click Fork in the header
3. Enter a name for your forked model
4. Your forked model appears in My Models

Forking copies the current HEAD checkpoint as your starting point.

Using Models via Gateway

All models on the platform are available through the HUD Gateway at inference.hud.ai:

from openai import AsyncOpenAI
import os

client = AsyncOpenAI(
    base_url="https://inference.hud.ai",
    api_key=os.environ["HUD_API_KEY"]
)

response = await client.chat.completions.create(
    model="claude-sonnet-4-5",  # Base model
    messages=[{"role": "user", "content": "Hello!"}]
)

For trained models, use the model’s API name from the Settings tab.

Next Steps

Gateway Docs

Learn the Gateway API in detail

A/B Testing

Compare models with variants