Large Language Models (LLMs) like GPT-4, Claude, and Gemini all have different API formats.

Inputs (requests) vary (prompt structure, temperature, top_p, etc.), and outputs (responses) also differ (metadata, reasoning, error handling). If you want to switch providers or build cross-LLM tooling, you need lots of custom adapters.

Open LLM Specification (OLLS) is an open, community-driven attempt to standardize both inputs & outputs for LLMs.

Example:

json

// Standardized input { "model": "gpt-4o", "task": "question_answering", "prompt": "What is the capital of France?", "parameters": { "temperature": 0.7 } }

// Standardized output { "content": "The capital of France is Paris.", "metadata": { "tokens_used": 123, "confidence": 0.95 } }

The goal is vendor-neutral interoperability—making it easier to:

Switch between LLM providers without rewriting code

Parse outputs consistently

Build universal middleware & tooling

We’re looking for contributors! Anyone can join the discussion, suggest fields, or share real-world needs.

Repo: https://github.com/julurisaichandu/open-llm-specification

Would love feedback from developers working with multiple LLMs. What fields/structures do you think should be mandatory vs optional?

Let’s make LLMs easier to work with across providers.