This is a very good entry into post training LLMs with RL. The whole recipe and data is open. Highly recommend!
Trained some terminal agents with friends!
Introducing Tmax, open RL terminal agent models. Under default settings and shorter length (65k) token budgets, tmax outperforms prior open work on terminal use. We are releasing all data+weights+rollouts publically!
Users praise Tmax's open RL Terminal Agent Models with full data and weights as a wonderful quick recipe for validating new post-training infrastructure.
No Digg Deeper questions have been answered for this story yet.
I've been using it as a quick means of validating my new post-training infrastructure this past week. Wonderful quick recipe to make sure the pieces are working beyond a simple hello world type RL run and fixing any bottlenecks in your rollouts and such.
This is a very good entry into post training LLMs with RL. The whole recipe and data is open. Highly recommend!
@Suhail Terminal-Bench 2.0 is one slice though, real long horizon agent work still favors closed frontier models by a wider margin.