Max Kirkby(@maxkirkby):I’m very excited that we @baselabs are today announcing our research efforts toward a common safety standard for open models, and our aim to build this into the infrastructure that serves these models at @baseten Our research will concern model alignment (and the training methods to meaningfully achieve this), interpretability and of course investigations where open models fail. As with our original promise, Base Labs plans to publish all the methods, code and evaluations behind our stack, such that the community can collectively build the safety standard for open models. We are initially excited to partner with @huggingface and @GoodfireAI to kickstart this, and invite others to join our journey!

2026.09.16 18:16

I’m very excited that we @baselabs are today announcing our research efforts toward a common safety standard for open models, and our aim to build this into the infrastructure that serves these models at @baseten Our research will concern model alignment (and the training methods to meaningfully achieve this), interpretability and of course investigations where open models fail. As with our original promise, Base Labs plans to publish all the methods, code and evaluations behind our stack, such that the community can collectively build the safety standard for open models. We are initially excited to partner with @huggingface and @GoodfireAI to kickstart this, and invite others to join our journey!