Essential AI

Open models with better pretraining efficiency rather than heavy post-training; uses the Muon optimizer. Rnj-1 is an 8B open model aimed at coding and agents.

Founded

2023

Headquarters

San Francisco, CA

Founders

Ashish Vaswani, Niki Parmar

Pushes against

RL-heavy post-training as the main source of gains.

Now

Released Rnj-1 on 2025-12-05. Funding on record is the 2023 Series A; later rounds not checked. No later release found in this pass.

Key people

• Ashish Vaswani (co-founder; co-author of Attention Is All You Need)

Funding

• December 2023

  Amount

  $56.5M Series A (about $65M total with an $8.3M seed)

  Lead

  March Capital (AMD, Franklin Venture Partners, Google, KB Investment, NVIDIA, Thrive Capital)

  Source venturebeat.com

Every funding round in the atlas

What it has shipped

No launch from this lab is in the atlas yet. These come from its profile.

• Rnj-1 (8B open model; reported 20.8 on SWE-bench Verified) 5 December 2025

People moves

Of the moves the atlas tracks, 1 joined and 0 left.

• Transformer paper authors (Vaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser, Polosukhin) 20 March 2024
  Google to Adept/Essential AI, Character.AI, Inceptive, Sakana AI, Cohere, OpenAI, Near
  Authors of 'Attention Is All You Need' (Google, 2017)

Sources

1. the-decoder.com/transformer-co-creator-vaswani-unveils-high-performance-rnj-1-coding-model
2. essential.ai/research/rnj-1
3. venturebeat.com/ai/essential-ai-emerges-from-stealth-with-backing-from-google-nvidia-and-a
4. pulse2.com/essential-ai-56-5-million-raised-to-enhance-automation/amp/

This profile was checked and corrected against its sources on 6 October 2026. How we check