Skip to content
Atlas by Comet Lab
Model layer

LLM providers, compared

These are the commercial model APIs most AI systems start with. The choice usually turns on fit more than on a benchmark: the cloud you already use, where your data may go, and what the work leans on. Whichever you pick, a gateway keeps switching cheap.

Learn the idea Which AI model is best? GPT, Claude, Gemini and the rest on Horizon

Open this layer in the explorer

At a glance

Every tool in this layer, with where it fits and the facts that usually decide it.

Tool Where it fits Licence Self-hostable Hosted API Maturity
OpenAI Frontier model API A safe default to start with when you want capable models, mature APIs, and the deepest pool of examples and support. Proprietary No Yes Mature
Amazon Bedrock Managed model platform The natural choice for AWS-committed organisations that want model access inside their existing account and governance. Proprietary No Yes Established
Anthropic Frontier model API A strong default when reasoning quality, long-context handling, or tool-use reliability are central to the work. Proprietary No Yes Established
Azure AI Foundry Managed model platform The default for Azure-committed enterprises wanting OpenAI and other models inside Azure governance and compliance. Proprietary No Yes Established
Cohere Enterprise model API A fit when retrieval and ranking quality matter, or when an enterprise wants private deployment and strong data-control terms. Proprietary No Yes Established
Google Gemini Frontier model API A natural fit for very long-context or multimodal work, and for teams already standardised on Google Cloud. Proprietary No Yes Established
Mistral AI Frontier model API A strong fit when European data residency or a non-US model vendor is a requirement, and for efficient, cost-aware models. Proprietary No Yes Established
xAI Grok Frontier model API A fit when you want another frontier-tier provider in the mix, particularly where access to real-time social data is useful. Proprietary No Yes Growing

How to choose

Start from your situation. Where a line says “our”, it comes from Comet Lab’s own builds.

  1. Starting a new build

    A safe default: capable models, mature APIs and the deepest pool of examples.

  2. Reasoning and tool use carry the work

    Claude is our frequent pick where reasoning, long context and tool-use reliability are central.

  3. Very long context, multimodal input, or a client on Google Cloud

    Some of the longest context windows on the market, and Vertex AI keeps it inside Google Cloud governance.

  4. Models must stay inside your cloud account

    Bedrock for AWS, Azure AI Foundry for Azure: model access inside the governance you already have.

  5. A European provider

    Our usual answer for EU data residency. It also publishes open weights, which makes a later move to self-hosting easier.

  6. Retrieval and ranking quality

    Models built for retrieval and ranking, with private deployment for enterprises.