@Suhail: This is a very good entry into post training LLMs with RL. The whole recipe and data is open. Highly recommend!
Summary
Hamish Ivison and team release Tmax, an open-source RL-trained terminal agent model that outperforms prior work under standard settings. All data, weights, and rollouts are publicly released.
View Cached Full Text
Cached at: 06/28/26, 12:00 PM
This is a very good entry into post training LLMs with RL. The whole recipe and data is open. Highly recommend!
Hamish Ivison @ ICML (@hamishivi): 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!
Similar Articles
@hamishivi: Trained some terminal agents with friends! Introducing Tmax, open RL terminal agent models. Under default settings and …
Introducing Tmax, open reinforcement learning terminal agent models that outperform prior open work on terminal use. All data, weights, and rollouts are being released publicly.
@adithya_s_k: [Must Read] I don't know how this flew under my radar. This is probably one of the best open source end-to-end post-tra…
An open-source end-to-end post-training recipe covering SFT, RL, and test-time scaling with models, datasets, and code.
RL Excursions during Pre-Training: Re-examining Policy Optimization for LLM training
Harvard researchers challenge the standard LLM training pipeline by showing RL can be effectively applied during pre-training rather than only after SFT, finding that data composition matters more than model scale, and proposing parallel averaging of RL and SFT objectives that outperforms sequential approaches while preserving general capabilities.
@maximelabonne: Such a cool collab, very happy to see this kind of recipe getting open-sourced Train LFM2.5-2.6B on all the harnesses!
Maximilien Labonne highlights an open-sourced guide to training models with RL inside real agent harnesses like Claude Code, Codex, and OpenCode, enabling training of any model (e.g., LFM2.5-2.6B) on any task set across harnesses.
Tmax: A simple recipe for terminal agents
Tmax introduces a simplified RL training recipe for terminal agents, achieving state-of-the-art performance with a 9B parameter model using a novel data generation taxonomy and an expanded open-source dataset.