# Guide Labs

> Guide Labs is a San Francisco AI research company building interpretable foundation models - large language and diffusion models that can reliably explain how they reach an answer, trace outputs back to their training data, and be steered without retraining. Founded in 2023 by machine-learning researcher Julius Adebayo and colleagues from the academic interpretability world, the company argues that the black-box status quo is a solvable engineering problem, not an unavoidable cost of scale. Its flagship open-source release, Steerling-8B, is pitched as the first large-scale inherently interpretable language model, and its Clarity platform packages interpretability tools for teams in regulated, high-stakes fields like lending, medicine, and drug discovery.

- **Founded:** 2023
- **Headquarters:** San Francisco, California, United States
- **Founders:** Julius Adebayo (Co-founder & CEO), Fulton Wang (Co-founder), Aya Abdelsalam Ismail (Chief Science Officer)
- **Team size:** Small - roughly 4-16 people (sources vary); founding team of researchers with PhDs from MIT, MILA, and UMD
- **Products:** Steerling-8B, Clarity, Interpretable foundation models, Concept-level steering & provenance
- **Notable:** Released Steerling-8B, described as the first large-scale inherently interpretable language model, open-sourced under Apache 2.0 (February 2026)., Reported that Steerling-8B reaches roughly 90% of the capability of larger, less interpretable models while using less training data., Raised a $9M seed round led by Initialized Capital (2024).

## Products & services

- **Steerling-8B** — An open-source, 8-billion-parameter interpretable language model that combines masked diffusion language modeling with a concept-decomposition architecture. Every token it produces can be traced to input context, training data, and human-understandable concepts. Released under Apache 2.0 on Hugging Face and GitHub.
- **Clarity** — Guide Labs' inherently interpretable AI platform (platform.guidelabs.ai), packaging the company's interpretability tooling so teams can inspect, debug, and steer models in production.
- **Interpretable foundation models** — A broader class of interpretable generative models developed by Guide Labs, including what it describes as the first interpretable generative diffusion model, aimed at high-accountability domains.
- **Concept-level steering & provenance** — Capabilities that let developers suppress or amplify specific concepts at inference time without retraining, and obtain training-data provenance for any generated chunk - enabling inference-time alignment via concept control.

## Achievements

- Released Steerling-8B, described as the first large-scale inherently interpretable language model, open-sourced under Apache 2.0 (February 2026).
- Reported that Steerling-8B reaches roughly 90% of the capability of larger, less interpretable models while using less training data.
- Raised a $9M seed round led by Initialized Capital (2024).
- Built what the company describes as the first interpretable generative diffusion model.
- Founder Julius Adebayo co-authored the influential 2018 'Sanity Checks for Saliency Maps' work questioning the reliability of existing model-explanation methods.
- Selected for Y Combinator (Winter 2024).

## Latest updates

- **2026-02** — Open-sourced Steerling-8B, an interpretable 8B LLM that traces every token to input context, training data, and human-understandable concepts; weights and code released under Apache 2.0.
- **2026-02** — Published research on aligning and controlling Steerling-8B via concept-level steering without retraining.
- **2024-11** — Closed a $9M seed round led by Initialized Capital to build large-scale interpretable language models.
- **2024-01** — Participated in Y Combinator's Winter 2024 batch.

## Links

- Website: https://www.guidelabs.ai/
- LinkedIn: https://www.linkedin.com/company/guide-labs/
- Twitter/X: https://x.com/guidelabsai
- GitHub: https://github.com/guidelabs

---

Profile page: https://yespress.io/guide-labs
Published by YesPress — https://yespress.io
Last updated: 2026-08-05