Foundation Model
What it means: A foundation model is a large, general-purpose AI model trained on vast amounts of text, code, or images. It serves as the base for many downstream applications. Why it matters: They’re trained once at great cost, then adapted for many uses, which makes them the “foundation” for a whole ecosystem of tools. GPT and Claude models are examples. How it shows up in Gloo: Gloo AI gives you foundation models from leading providers, including OpenAI, Anthropic, and Google, through one API. You can add Gloo’s guardrails and values alignment with the guarded endpoints, or ground answers in your own content.Frontier Model
What it means: A frontier model is one of the newest and most capable AI models available. Why it matters: These models push the limits of what AI can do. They’re usually the most expensive, and they need careful alignment. How it shows up in Gloo: Gloo AI’s catalog includes current frontier models, such as the latest Claude, GPT, and Gemini models. You can pick one by name, or turn on auto-routing on Guarded Completions to let Gloo choose a model for each request based on what it needs.Open-Weight Model
What it means: An open-weight model’s trained parameters (its “weights”) are published. Anyone can download, run, or fine-tune it. Why it matters: Open weights make models easier to inspect, experiment with, and run in many places. Meta’s Llama, Mistral, DeepSeek, and gpt-oss are examples. How it shows up in Gloo: Gloo AI’s catalog includes open-weight models such as DeepSeek, GLM, and gpt-oss, alongside closed models. They use the same API, and the guarded endpoints apply the same guardrails to them.Closed Model
What it means: Closed models are proprietary, with their weights, training data, and internal design kept private. Why it matters: Closed models, like OpenAI’s GPT models, are often highly capable, but the public can’t inspect or modify them. You use them through the provider’s API. How it shows up in Gloo: Gloo AI gives you closed models from OpenAI, Anthropic, and Google through the same API as every other model. You can call them directly, through the guarded endpoints, or with grounding in your own content.Multilingual Model
What it means: These models understand and generate content in multiple languages. Why it matters: Multilingual support makes models globally useful and more inclusive. However, performance can vary across languages. How it shows up in Gloo: Most models on Gloo AI can read and write many languages. How well they do varies by model and language, so test with your own content and audience.Instruction-Tuned Model
What it means: This is a model fine-tuned to follow human instructions, usually on examples of prompt-response pairs. Why it matters: Instruction tuning helps models understand tasks written as plain-language instructions, which makes them much easier to work with. How it shows up in Gloo: The chat models on Gloo AI are instruction-tuned, which is why a clear prompt or system prompt works so well with them. You add your own voice and rules with a system prompt, and the guarded endpoints add guardrails and values alignment.What’s next
When you build with Gloo’s API, Supported Models lists every model by type, and Choosing, Pinning and Updating Models helps you pick one for each task.What Kind of Hardware Powers AI, and Why Does It Matter?
What hardware does AI depend on, and how do specialized chips and infrastructure affect cost and performance?

