ToolSpace-744B-A40B
Code · Evaluation · 8B · 35B-A3B · 744B-A40B
ToolSpace-744B-A40B is the GLM-5.2-based mixture-of-experts agent evaluated in the ToolSpace paper, with 744B total and 40B active parameters. This repository distributes the official Atria-Dawn full-release checkpoint.
ToolSpace constructs a tool-composition graph from real execution traces, capturing data, state, and retry dependencies. It synthesizes cross-environment tasks through state-aware walks with rollback and keeps trajectories that pass execution-based verification.
Paper: ToolSpace: Scaling Tool-Centric Environment Synthesis for Long-Horizon Agents.
https://huggingface.co/papers/2611.05036
Model information
| Item | Details |
|---|---|
| Backbone | GLM-5.2 |
| Architecture | MoE, 744B total / 40B active parameters |
| Checkpoint | Official Atria-Dawn full release |
| Context length | 512K, following the official release |
| Weight format | BF16 Safetensors |
Results
Results from Table 1 of the ToolSpace paper. Avg. is the unweighted mean over the seven benchmarks.
| Model | AutomationBench | SkillsBench | JobBench | τ³-Bench Banking | BFCL V4 | Harvey LAB | OfficeQA Pro | Avg. |
|---|---|---|---|---|---|---|---|---|
| GLM-5.2 | 29.3 | 56.7 | 43.4 | 37.1 | 75.6 | 13.3 | 56.9 | 44.6 |
| ToolSpace-744B-A40B | 54.7 | 67.8 | 57.5 | 43.3 | 78.4 | 27.5 | 67.1 | 56.6 |
Usage
Follow the official Atria-Dawn deployment guide for distributed SGLang or vLLM serving, using dongguanting/ToolSpace-744B-A40B as the model path. The BF16 checkpoint is approximately 1.51 TB.
Use the tokenizer, configuration, and chat template included in this repository.
Evaluation commands and benchmark configurations are available in the ToolSpace evaluation directory.