MIT Open Source antidrift

anymodel

LLM router with unified batch support. 11 providers, 323 models. Write once, switch providers without changing code.

Node.js · Python · Go

11 providers

323 models

50% batch cost savings

0 required config

$ npm install @probeo/anymodel

$ pip install anymodel-py

$ go get github.com/probeo-io/anymodel-go

What it does

anymodel wraps 11 LLM providers behind one API. Change provider in your config and the rest of your code stays the same. It supports native batch for OpenAI, Anthropic, and Google — submitting jobs to the providers' batch endpoints directly, not faking it with sequential calls. That's where the 50% cost reduction comes from.

Fallback routing retries with the next provider in your priority list if one is rate-limited or down. Streaming and tool calling work across all providers.

Usage

Node.js

import { anymodel } from '@probeo/anymodel';

const client = anymodel({ provider: 'anthropic', model: 'claude-sonnet-4-6' });

// Single call
const res = await client.complete({ prompt: 'Summarize this doc.' });

// Batch (50% cost savings for OpenAI/Anthropic/Google)
const batch = await client.batch([
  { prompt: 'Analyze page 1.' },
  { prompt: 'Analyze page 2.' },
  { prompt: 'Analyze page 3.' },
]);

// Streaming
for await (const chunk of client.stream({ prompt: 'Write a report.' })) {
  process.stdout.write(chunk.text);
}

Fallback routing

const client = anymodel({
  providers: [
    { provider: 'anthropic', model: 'claude-sonnet-4-6' },
    { provider: 'openai',    model: 'gpt-4o' },
    { provider: 'google',    model: 'gemini-2.5-pro' },
  ],
});
// Tries providers in order on rate limit or error

Providers

OpenAI Anthropic Google Mistral Groq DeepSeek xAI Together Fireworks Perplexity Ollama

Used in Probeo

anymodel was extracted from Probeo, where it runs in production across multiple pipeline stages.

View on GitHub → ← All tools