research-llm-apis

Overview

research-llm-apis is a repository created to document and compare the HTTP APIs offered by different large language model providers. It appears to be designed as a hands-on research artifact, with scripts and captured outputs covering both streaming and non-streaming behaviors across providers. The stated purpose is to inform abstraction-layer decisions for developers building software that needs to work across multiple model vendors.

For AI Product Managers, this matters because provider interoperability is rarely just a model-quality question—it is also an API design, reliability, and product architecture question. A resource like research-llm-apis helps teams understand where providers differ in request formats, response structures, streaming semantics, and operational behavior. That makes it useful for planning multi-provider strategies, reducing switching costs, and making more informed roadmap decisions around platform flexibility.

Key Developments

• 2026-04-05: Simon Willison's research-llm-apis repository was highlighted as a new effort capturing research into multiple LLM providers' HTTP APIs.
• 2026-04-05: The repository was described as supporting a major planned change to the abstraction layer in the LLM Python library, with scripts and saved outputs for both streaming and non-streaming modes.

Relevance to AI PMs

• Evaluate multi-provider risk earlier: AI PMs can use the repository's research framing to identify where API incompatibilities may create engineering overhead, vendor lock-in, or feature gaps before committing to a provider strategy.
• Inform abstraction-layer requirements: If your product may support multiple LLM vendors, this repository highlights the importance of comparing streaming behavior, response formats, and endpoint capabilities when defining platform requirements.
• Improve roadmap and migration planning: Research into provider API differences can help PMs scope the true cost of switching providers, adding fallback providers, or building a unified internal API for product teams.

Related

• Simon Willison: The repository is associated with Simon Willison, who highlighted it as a research vehicle for understanding provider API differences.
• llm-python-library: research-llm-apis directly connects to the LLM Python library because its findings are intended to inform a major change to that library's abstraction layer.