Mozilla launches unified LLM tool
Mozilla has released *any-llm*, a tool that enables provider-agnostic integration across APIs from OpenAI, Anthropic, and Mistral, as well as locally-run models. The tool supports integrations with JupyterLiteAI and LangChain, pointing toward a trend of using API abstraction layers to future-proof AI engineering pipelines and simplify experimentation.
- The version 1.0 of *any-llm* was released on November 4, 2025, moving the tool from an initial release to a production-ready library with a stable API surface and async-first APIs for high-throughput use cases. - A key technical differentiator is its use of official provider SDKs, which contrasts with other popular tools that re-implement provider interfaces and risk compatibility issues. - The tool can programmatically query supported models from each provider using a "List Models" API, allowing for dynamic model selection and validation within an application at runtime. - Beyond the initial Python library, Mozilla.ai released "any-llm-go" on February 10, 2026, extending the same provider-agnostic principles to the Go programming language. - This tool is part of a larger strategic push into AI by Mozilla, whose new CEO, Anthony Enzor-DeMeo, laid out a plan in December 2025 to transform Firefox into a "modern AI browser." - Provider-agnostic tools are critical for MLOps as they mitigate risks like vendor lock-in, service outages (e.g., the ChatGPT outage on January 23), and allow for cost optimization by selecting the best-performing model for a specific task. - Future plans for the library include adding support for native batch completions and deeper integrations with other tools in Mozilla's ecosystem, such as *any-guardrail* for security and *any-agent*.
