Tmax: A simple recipe for terminal agents
cs.CL
Submitted: 2026-06-22
Updated: 2026-11-02
Comments: preprint
Code: https://github.com/hamishivi/tmax
License: http://creativecommons.org/licenses/by/4.0/
The gist: Terminal-using agents have quickly become a popular and important downstream application of language models (LMs).
Terminology
Abstract
Terminal-using agents have quickly become a popular and important downstream application of language models (LMs). Despite their prevalence, relatively little academic work has examined RL-based training of these models, due to difficult benchmarks, a lack of data, and a lack of simple baseline recipes. We present TMax, an open RL recipe for terminal agents that brings small open models closer to the frontier. While simple, our recipe achieves 33% on Terminal-Bench 2.0 with only 9B parameters, outperforming much larger models from prior work. Our recipe first involves generating data using a novel taxonomy, combining difficulty control, personas, and verifier diversification, which allows us to cheaply generate large amounts of terminal environments for RL and SFT training. We then train open-weight models using RL with our data, using a simple, outcome-only recipe, with small adjustments to improve stability. Our recipe works for models from 2B to 27B and across model families, dominating the Pareto curve for open terminal-agent performance at smaller sizes. We release our data, models, and code as a strong baseline for future open academic work on terminal agents at https://github.com/hamishivi/tmax.
Sources
- SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agents
- SWE-Universe: Scale Real-World Verifiable Environments to Millions
- Computer Environments Elicit General Agentic Intelligence in LLMs
- Composer 2 Technical Report
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Endless Terminals: Scaling RL Environments for Terminal Agents
- Scaling Synthetic Data Creation with 1,000,000,000 Personas
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training
- CLI-Gym: Scalable CLI Task Generation via Agentic Environment Inversion
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
- MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
- Olmo 3
- Training Software Engineering Agents and Verifiers with SWE-Gym
- On Data Engineering for Scaling LLM Terminal Capabilities
- Rethinking the Trust Region in LLM Reinforcement Learning
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- SERA: Soft-Verified Efficient Repository Agents
- Kimi k1.5: Scaling Reinforcement Learning with LLMs
- LLaMA: Open and Efficient Foundation Language Models
- Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data