tmax-8b-v1.1

<p align="center"> 💻 <a href="https://github.com/hamishivi/tmax">Code</a> · 🤗 <a href="https://huggingface.co/collections/TMaxxx/tmax-v11">Models &amp; Data</a> · 📜 <a href="https://arxiv.org/abs/2606.23321">Paper</a> · 📓 <a href="https://wai-org.com/blog/tmax/">Blog</a> </p>

TMax v1.1 8B

This terminal-agent model was trained using DPPO on allenai/tmax-sft-8b with the TMax-13K v1.1 task set. It starts from TMax SFT 8B, a Qwen3-8B SFT checkpoint; this release is the subsequent RL model.

This model is part of the TMax v1.1 collection. The original TMax recipe is described in our paper.

Checkpoints

This branch contains step 500. The main branch contains step 500, the highest-scoring available checkpoint in the completed TB-Lite sweep. Checkpoint selection uses TB-Lite, not Terminal-Bench 2.1.

Checkpoints 100, 200, 300, 400, 500 are provided as branches named step-0100, step-0200, etc.

Evaluation Results

The selected checkpoint scores 6.44 ± 1.00% on Terminal-Bench 2.1 (88 tasks, three trials per task; configure-git-webserver excluded).

These v1.1 scores use the fixed Vanillux0.2.3 harness with Sandfleet through Harbor, three trials per task (96 TB-Lite tasks). Scores and standard errors are taken from the completed evaluation sweep updated on October 5, 2026. The harness differs from the original release; use the original model cards for v1.0 results. The per-checkpoint results and pinned source revisions are included in release_manifest.json.

Model Details

Model Description

• Language: English
• License: Apache 2.0
• Finetuned from: allenai/tmax-sft-8b
• Task set: TMaxxx/tmax-13k-v1.1
• Training method: DPPO reinforcement learning

Use

Serve the model with vLLM and use the terminal-agent harness in our codebase:

Use a Linux NVIDIA GPU environment with a CUDA toolkit matching PyTorch (CUDA_HOME set and nvcc on PATH; see the vLLM installation guide). The tested environment uses Python 3.12, PyTorch 2.13.0, and CUDA 13.0.

uvx --python 3.12 vllm==0.32.0 serve TMaxxx/tmax-8b-v1.1 \
  --served-model-name tmax-v1.1 \
  --enable-auto-tool-choice \
  --tool-call-parser hermes \
  --port 8008 \
  --max-model-len 40960 \
  --tensor-parallel-size 1

The example uses one GPU. Increase --tensor-parallel-size or reduce --max-model-len to fit your available GPU memory.

Qwen3-8B uses the Hermes tool parser and has a 40,960-token context limit.

Training Details

• Released training step: 500
• Main checkpoint: 500, selected on TB-Lite
• Available milestone checkpoints: 100, 200, 300, 400, 500

The source checkpoint is TMaxxx/tmax-8b-v1.1 at step 500. Original checkpoint files are preserved, including campaign_provenance.json where supplied. See release_manifest.json for the immutable source commits and evaluation results. See the training launch scripts and original provenance for run-specific settings.

License

This model is licensed under Apache 2.0.

Citation

If you use our model or data, please cite our paper:

@misc{ivison2026tmaxsimplerecipeterminal,
      title={Tmax: A simple recipe for terminal agents}, 
      author={Hamish Ivison and Junjie Oscar Yin and Rulin Shao and Teng Xiao and Nathan Lambert and Hannaneh Hajishirzi},
      year={2026},
      eprint={2606.23321},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2606.23321}, 
}