AI News HubLIVE
In-site rewrite3 min read

Neuronpedia, an open source platform for AI interpretability

Neuronpedia is an open source interpretability platform that enables users to explore, visualize, and steer the internal workings of AI models. It features tools like HeadVis, Natural Language Autoencoders, Circuit Tracer, and steerable activations, supporting over 50 million latent vectors across numerous models. Created by Johnny Lin, it is backed by organizations including Anthropic and Google DeepMind.

SourceHacker News AIAuthor: rzk

Neuronpedia

© Neuronpedia 2026

Privacy & TermsBlogGitHubSlackTwitterContact

Neuronpedia

Jacobian LensNEW

Natural Language

Autoencoders

NEWAssistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact

Neuronpedia is an open source interpretability platform.

Explore, visualize, and steer the internals of AI models.

Gurnee et al.

Jacobian Lens

Revealing a Global Workspace in Language Models

Fraser-Taliente, Kantamneni, Ong et al.

Natural Language Autoencoders

Translate a Model's Internal Thoughts Into Text

Lu et al.

Assistant Axis

Monitor & Stabilize the Character of an LLM

Multi-Org

Circuit Tracer

Trace a Model's Internal Reasoning Steps

Google Deepmind

Gemma Scope 2

SAEs and Transcoders for Gemma 3

Latest Update

June 2026

HeadVis, plus NLA Contributions and more SAEs

Explore 36,000+ Attention Heads Across 37 Models

Newsletter

No spam, unsubscribe anytime.

Releases and Models

Browse five+ terabytes of activations, explanations, and metadata. Neuronpedia supports probes, latents/features, custom vectors, concepts, and more.

Releases

Featured

HeadVis: Browse & Investigate Attention Heads

Luger, Kamath, et al., Anthropic

Featured

Natural Language Autoencoders

Fraser-Taliente, Kantamneni, Ong et al., Anthropic

Featured

Circuit Tracing with Interpretable Attention

OpenMOSS Team, Fudan University

Featured

Assistant Axis: Situating and Stabilizing the Character of an LLM

Lu et al, Anthropic Fellows

Featured

Gemma Scope 2: Suite of SAEs and Transcoders for Gemma 3

Language Model Interpretability Team, Google DeepMind

Featured

Circuit Tracer: Tracing the Internal Reasoning Steps of a Model

Hanna & Piotrowski, Anthropic Fellows

Featured

Gemma Scope - Exploring the Inner Workings of Gemma 2

Language Model Interpretability Team, Google DeepMind

Featured

SAE Bench

Adam Karvonen · Can Rager

Gemma 4 SAEs

David Chanin

Qwen 3.5 SAEs

David Chanin

Olmo 3 7B SAE

David Chanin

Olmo 3 32B SAE

Bartosz Cywinski

Qwen Scope

Alibaba

Qwen3 SAEs

Adam Karvonen

Qwen2.5 7B IT SAE

David Chanin

Beta

Weight-Sparse Transformers Have Interpretable Circuits

OpenAI

Temporal Feature Analysis

Lubana, Rager, Hindupur, et al.

gpt-oss BatchTopK SAEs

Andy Arditi

Finding Misaligned Persona Features in Open-Weight Models

Andy Arditi

Circuit Tracer Transcoders

Hanna & Piotrowski

A Bunch of Matryoshka SAEs

David Chanin

Llama 3.3 70B Instruct SAE

Goodfire

Llama Scope R1: SAEs for DeepSeek-R1-Distill-Llama-8B

OpenMOSS Team, Fudan University

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

pyvene.ai, The Stanford NLP Group

Llama Scope: SAEs for Llama-3.1-8B

OpenMOSS Team, Fudan University

Feature Splitting for GPT2-Small

Joseph Bloom

Multi TopK SAE for Llama3.1-8B

EleutherAI

Sparse Autoencoder for GPT2-Small - v5

OpenAI

Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning

Apollo Research · Jordan Taylor

Transcoders Enable Fine-Grained Interpretable Circuit Analysis for Language Models

Jacob Dunefsky · Philippe Chlenski

Sparse Autoencoders for Pythia-70M-Deduped

Under Peer Review

Attention SAE Research Paper

Under Peer Review

Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small

Joseph Bloom

Models

GEMMA-4-31B

Gemma-4-31BGoogle Deepmind

QWEN3.5-4B

Qwen3.5-4BAlibaba

GEMMA-4-E2B

Gemma-4-E2BGoogle Deepmind

GEMMA-4-E4B

Gemma-4-E2BGoogle Deepmind

QWEN3.5-0.8B

Qwen3.5-0.8BAlibaba

OLMO-3-1025-7B

Olmo-3-1025-7BAllen AI

OLMO-3-1125-32B

Olmo-3-1125-32BAllen AI

QWEN3.6-27B

Qwen3.6-27BAlibaba

QWEN3.5-27B

Qwen3.5-27BAlibaba

QWEN3-32B

Qwen3-32BAlibaba

QWEN3.5-9B-PT

Qwen3.5-9B BaseAlibaba

QWEN3.5-2B-PT

Qwen3.5-2B BaseAlibaba

QWEN3-14B

Qwen3-14BAlibaba

QWEN3-8B

Qwen3-8BAlibaba

CIRCUITGPT-PYTHON

CircuitGPT-PythonOpenAI

GEMMA-3-27B

Gemma-3-27BGoogle Deepmind

GEMMA-3-12B

Gemma-3-12BGoogle Deepmind

GEMMA-3-270M-IT

Gemma-3-270M-ITGoogle Deepmind

GEMMA-3-1B-IT

Gemma-3-1B-ITGoogle Deepmind

GEMMA-3-4B-IT

Gemma-3-4B-ITGoogle Deepmind

GEMMA-3-12B-IT

Gemma-3-12B-ITGoogle Deepmind

GEMMA-3-27B-IT

Gemma-3-27B-ITGoogle Deepmind

GEMMA-3-270M

Gemma-3-270MGoogle Deepmind

GEMMA-3-4B

Gemma-3-4BGoogle Deepmind

GEMMA-3-1B

Gemma-3-1BGoogle Deepmind

GEMMA-2-27B

Gemma-2-27BGoogle Deepmind

GPT-OSS-20B

GPT-OSS-20BOpenAI

LLAMA3.1-8B-IT

Llama3.1-8B-ITMeta

QWEN2.5-7B-IT

Qwen2.5-7B-ITAlibaba

QWEN3-1.7B

Qwen3-1.7BAlibaba

QWEN3-4B

Qwen3-4BAlibaba

LLAMA3.3-70B-IT

Llama3.3-70B-ITMeta

GEMMA-2-2B-IT

Gemma-2-2B-ITGoogle Deepmind

GEMMA-2-9B-IT

Gemma-2-9B-ITGoogle Deepmind

LLAMA3.1-8B

Llama3.1-8B (Base)Meta

GEMMA-2-2B

Gemma-2-2BGoogle Deepmind

GEMMA-2-9B

Gemma-2-9BGoogle Deepmind

P70M-D

Pythia-70M-DedupedEleutherAI

GPT2-SMALL

GPT2-SmallOpenAI

Jump To

Jump to Source/SAE

Jump to Feature

INDEX

Jump to Random

Graph

Visualize and trace the internal reasoning steps of a model with custom prompts, pioneered by Anthropic's circuit tracing papers.

Steer

Modify model behavior by steering its activations using latents or custom vectors. Steering supports instruct (chat) and reasoning models, and has fully customizable temperature, strength, seed, etc.

Search

Search over 50,000,000 latents/vectors, either by semantic similarity to explanation text, or by running custom text via inference through a model to find top matches.

Search via Inference

Run Example Search

API + Libraries

Neuronpedia hosts the world's first interpretability API (March 2024) - and all functionality is available by API or Python/TypeScript libraries. Most endpoints have an OpenAPI spec and interactive docs.

Inspect

Go in depth on each probe/latent/feature with top activations, top logits, activation density, and live inference testing. All dashboards have unique links, can be compiled into sharable lists, and supports IFrame embedding, as demonstrated here.

Who We Are

Neuronpedia was created by Johnny Lin, an ex-Apple engineer who previously founded a privacy startup. Neuronpedia is supported by Decode Research, Open Philanthropy, the Long Term Future Fund, AISTOF, Anthropic, Manifund, and others.

Get Involved

Citation

@misc{neuronpedia, title = {Neuronpedia: Interactive Reference and Tooling for Analyzing Neural Networks}, year = {2023}, note = {Software available from neuronpedia.org}, url = {https://www.neuronpedia.org}, author = {Lin, Johnny} }