Supported models

Poolside provides proprietary AI models for software engineering. You deploy and use these models entirely within your environment, so your code, prompts, and outputs stay under your control. Supported models are Poolside-managed model families that Poolside validates, ships, upgrades, patches, and helps operate as part of your Poolside platform. External chat model providers are compatibility targets. Poolside supports connecting to these providers from your deployment, but the third-party provider is responsible for model behavior, model availability, and provider-specific changes. Poolside models can work in one or more modes:

• Agentic: Plans and executes multi-step tasks, iterating and applying code changes, refactors, and suggestions.
• Chat: Answers questions and provides explanations in a conversational interface.
• Completion: Provides inline and multi-line code suggestions as you type in your editor.

​

Supported models

For hardware sizing guidance and concurrent-agent capacity by model, see Capacity planning.

​

External chat model providers

Poolside connects to chat models through any provider that exposes an OpenAI-compatible chat completions API. This section lists providers that have been tested with Poolside and their current compatibility status.

​

Tested providers

Although each of the providers below has been tested with a subset of models available, given the large number of possible model-provider combinations, not every model-provider combination may work as expected. The following providers are known to work with Poolside:

• OpenAI Chat Completions (not Responses)
• Amazon Bedrock Mantle
• Microsoft Azure Foundry
• Google Vertex
• LiteLLM
• Ollama
• OpenRouter
• Hugging Face Inference Providers
• Baseten Model APIs

​

Known issues

The following providers have known compatibility issues. You may encounter unexpected behavior when using these providers with Poolside:

• Fireworks

To configure an external model, see Create a model provider.