Update: Sentience now powered by gpt-oss
Sentience has officially evolved from a Gemma-based prototype into a full autonomous multimodal cognition engine running on OpenAI’s gpt-oss models.
Key Highlights
- gpt-oss-20b integration via Ollama on Apple Silicon for fully offline reasoning
- gpt-oss-120b integration via vLLM on GPU servers for high-performance cognition
- Harmony compliance built into the runtime for stable, structured outputs
- Modular planner backend: switch between Gemma, gpt-oss, or OpenAI endpoints with a single flag
- Improved test mode: run synthetic video/audio inputs to validate without a webcam or mic
What’s Next
The next milestone is wiring Sentience into robotics backends, extending cognition from observation to real-world action.
Log in or sign up for Devpost to join the conversation.