Mozilla.ai, supported by the Mozilla Foundation, has officially released any-llm v1.0, an open-source Python library designed to provide developers with a unified interface for interacting with multiple large language model (LLM) providers. The library allows developers to switch seamlessly between cloud-based and local models without rewriting code or overhauling their existing stacks. This flexibility reduces boilerplate code, minimizes integration challenges, and gives developers the freedom to choose the best model for their use case, according to Nathan Brake, machine learning engineer at Mozilla.ai.
The initial version of any-llm was introduced on July 24, but the 1.0 release, made available on November 4 via GitHub, offers a stable API, async-first functionality, and reusable client connections optimized for high-throughput and streaming applications. Brake emphasized that the release includes clear deprecation and experimental notices, helping developers anticipate API changes and avoid surprises during integration.
Key features of any-llm v1.0 include improved test coverage for reliability, a Responses API, a List Models API for querying available models per provider, and reusable client connections for better performance. Additionally, the library standardizes reasoning outputs across all models, so developers can access consistent results regardless of which provider they select. The library also includes an auto-updating provider compatibility matrix, making it easier to track supported features across different LLMs.
Looking ahead, Mozilla.ai plans to enhance any-llm with native batch completions, support for additional model providers, and deeper integration with its broader “any-suite” libraries, such as any-guardrail, any-agent, and MCPD. By simplifying multi-provider LLM development, any-llm v1.0 aims to empower developers to build more robust, flexible, and scalable AI applications without being locked into a single ecosystem.