# Essential AI

> Essential AI is a San Francisco AI research company founded in 2023 by Ashish Vaswani and Niki
> Parmar, co-authors of "Attention Is All You Need." It builds open-source models, datasets, and
> training infrastructure — most notably the Apache 2.0 Rnj-1 family of 8B open-weight base and
> instruction-tuned models (32K context, tuned for code, STEM reasoning, and agentic tool use) and
> the Essential-Web v1.0 web dataset. Essential AI publishes NO first-party HTTP API, developer
> portal, OpenAPI/AsyncAPI contract, or MCP server; its public interface is open model weights,
> datasets, and research code.

## APIs

- No public API. Essential AI operates no first-party HTTP API. Probes on 2026-08-04 of
  essential.ai, www.essential.ai, api.essential.ai and app.essential.ai found no OpenAPI, Swagger,
  GraphQL, AsyncAPI, MCP, or A2A agent card. api.essential.ai and app.essential.ai resolve via
  Cloudflare but returned 522/526 with no served content.

## Models

- [Rnj-1](https://huggingface.co/EssentialAI/rnj-1): 8B dense open-weight base model, 32K context, Apache 2.0.
- [Rnj-1 Instruct](https://huggingface.co/EssentialAI/rnj-1-instruct): instruction-tuned Rnj-1 for code, math/STEM reasoning, and agentic tool use.
- [Rnj-1.5 Instruct](https://huggingface.co/EssentialAI/rnj-1.5-instruct): current instruct release.
- [Rnj-1 Instruct GGUF](https://huggingface.co/EssentialAI/rnj-1-instruct-GGUF): quantized builds for llama.cpp, Ollama, LM Studio.
- [EAI Distill 0.5B](https://huggingface.co/EssentialAI/eai-distill-0.5b): small taxonomy/classification model.

## Datasets

- [Essential-Web v1.0](https://huggingface.co/datasets/EssentialAI/essential-web-v1.0): document-level web dataset organized by a twelve-category taxonomy (ODC-BY).
- [eai-taxonomy-code-w-dclm](https://huggingface.co/datasets/EssentialAI/eai-taxonomy-code-w-dclm), [eai-taxonomy-stem-w-dclm](https://huggingface.co/datasets/EssentialAI/eai-taxonomy-stem-w-dclm), [eai-taxonomy-med-w-dclm](https://huggingface.co/datasets/EssentialAI/eai-taxonomy-med-w-dclm), [eai-taxonomy-math-w-fm](https://huggingface.co/datasets/EssentialAI/eai-taxonomy-math-w-fm): domain-filtered corpora.
- [Rnj-1 evals](https://huggingface.co/datasets/EssentialAI/rnj-1-instruct-evals): published evaluation outputs.

## Source code

- [Essential-AI on GitHub](https://github.com/Essential-AI): research org.
- [eai-taxonomy](https://github.com/Essential-AI/eai-taxonomy): the Essential-Web taxonomy classifier.
- [reflection](https://github.com/Essential-AI/reflection): reflection research code (MIT).
- [maxtext-external](https://github.com/Essential-AI/maxtext-external): JAX LLM training infrastructure (Apache 2.0).
- [homebrew-essentialai](https://github.com/Essential-AI/homebrew-essentialai): Homebrew tap for the RNJ macOS app.

## How to run the models

- Homebrew (macOS desktop app): `brew install --cask Essential-AI/essentialai/rnj` — RNJ.app, requires Ollama and macOS Ventura or later.
- [Ollama](https://ollama.com/library/rnj-1): `ollama run rnj-1` (requires Ollama 0.13.3+).
- [LM Studio](https://lmstudio.ai/models/essentialai/rnj-1).
- Third-party hosted inference: [Together AI](https://www.together.ai/models/rnj-1-instruct), [OpenRouter](https://openrouter.ai/essentialai/rnj-1-instruct). These are the inference providers' APIs, not Essential AI's.

## Research

- [Research index](https://www.essential.ai/research)
- [Rnj-1](https://www.essential.ai/research/rnj-1)
- [Evaluation](https://www.essential.ai/research/eval)
- [Grokking](https://www.essential.ai/research/grokking)
- [Infrastructure](https://www.essential.ai/research/infra)
- [Optimizer (Muon)](https://www.essential.ai/research/optimizer)

## Company

- [Home](https://www.essential.ai/)
- [About](https://www.essential.ai/about)
- [News](https://www.essential.ai/news)
- [Careers](https://www.essential.ai/careers)
- [Discord community](https://discord.gg/VPEqUNg6tR)
- [X](https://x.com/essential_ai), [LinkedIn](https://www.linkedin.com/company/essentialai)
- Contact: info@essential.ai

## Optional

- Independent API Evangelist profile: https://github.com/api-evangelist/essential-ai
- Generated by the API Evangelist enrichment pipeline on 2026-08-04 from public sources. Essential AI does not publish an llms.txt of its own (/llms.txt returned 404 on both essential.ai and www.essential.ai).