Guide Labs
ActiveAI-nativeFrontier models and agents you can learn from, audit, and steer.
What it does
Guide Labs builds frontier AI models and agents that you can learn from, audit, and steer. Our first model, Steerling-8B, is the first and largest inherently interpretable model: it can trace any output to training data, concepts inside the model, and the input, all without sacrificing performance.
Its site says now · captured 2026-08-23
Guide Labs
We are building a new class of interpretable AI systems and foundation models that humans can reliably debug, trust, and understand.
Next to it in AI infra and compute · Interpretable frontier models
- LinumY Combinator W23
type text, dream video
- Paper InstrumentsY Combinator X25
Open-source frontier models for knowledge work
- Cactus ComputeY Combinator S25
Tiny Edge AI For Tiny Devices
- Stellon LabsY Combinator S25
Building tiny frontier AI models that run on edge devices
- Uplift AIY Combinator S25
Foundational Voice Models for regional languages
- Kalpa LabsY Combinator F25
Scaling Generalist Speech models
- Miso LabsY Combinator X26
Emotive foundation voice models
- FastinoPlug and Play PnP 2026
Fastino provides task-specific language models for efficient and accurate AI development.