LLM Usage

For Code Reviews and the Coding Assistant Sourcery uses different large language models (LLMs) to help generate our reviews and code suggestions.

We are frequently evaluating new models and looking to determine what models are performing the best for specific tasks.

The models we currently use are:

• Anthropic Sonnet 3.5 (zero day retention)
• Anthropic Claude 2 (limited use)
• OpenAI GPT4-o
• Gemini 1.5

If you require zero retention set up, would like to use your own model deployments, or are looking for self hosted versions of Sourcery please contact us.