General Agents Benchmark

Workspace-Bench-Lite

Workspace-Bench-Lite - Lite split of Workspace-Bench 1.0, benchmarking AI agents on long-horizon workspace and large-scale file tasks. Scores are measured over the Workspace-Bench-Lite dataset (https://huggingface.co/datasets/Workspace-Bench/Workspace-Bench-Lite); project: https://workspace-bench.github.io/; paper: https://arxiv.org/abs/2605.03596.

Category

General Agents

Score type

percent

Scale

100%

higher is better

Models tested

3

What is Workspace-Bench-Lite?

Workspace-Bench-Lite is a benchmark for how well a model acts as a general-purpose agent. The official description says: "Workspace-Bench-Lite - Lite split of Workspace-Bench 1.0, benchmarking AI agents on long-horizon workspace and large-scale file tasks". Below we describe in simple words what benchmarks of this type generally examine.

What does it test?

Workspace-Bench-Lite measures how well a model acts as a general-purpose agent. In detail, it checks:

  • solving long, multi-step tasks with many intermediate decisions
  • using tools correctly (browsers, terminals, APIs, calendars, other apps)
  • adapting when something unexpected happens mid-task
  • staying on goal over long task horizons instead of wandering off

How to read the score

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

Official methodology note: Workspace-Bench-Lite - Lite split of Workspace-Bench 1.0, benchmarking AI agents on long-horizon workspace and large-scale file tasks. Scores are measured over the Workspace-Bench-Lite dataset (https://huggingface.co/datasets/Workspace-Bench/Workspace-Bench-Lite); project: https://workspace-bench.github.io/; paper: https://arxiv.org/abs/2605.03596.

Frequently asked questions

What is the Workspace-Bench-Lite benchmark?

Workspace-Bench-Lite is a benchmark for how well a model acts as a general-purpose agent. The official description says: 'Workspace-Bench-Lite - Lite split of Workspace-Bench 1.0, benchmarking AI agents on long-horizon workspace and large-scale file tasks'. Below we describe in simple words what benchmarks of this type generally examine.

What does Workspace-Bench-Lite test?

Workspace-Bench-Lite mainly measures how well a model acts as a general-purpose agent. Concretely, it checks: solving long, multi-step tasks with many intermediate decisions; using tools correctly (browsers, terminals, APIs, calendars, other apps); adapting when something unexpected happens mid-task.

How is the Workspace-Bench-Lite score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get Workspace-Bench-Lite data and scores?

The official source is https://huggingface.co/datasets/Workspace-Bench/Workspace-Bench-Lite. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/workspace-bench-lite/scores) or pull it directly from the Open LLM Radar API with any HTTP client.

Model scores 3 models tested

Reset
# Model Score Date Source Verified
1 Sage-S2.1-Flash
Shanghai AI Laboratory
67.1%
— Report Verified
2 DeepSeek-V4.1-Flash-1017
DeepSeek
66.8%
— Report Unverified
3 GLM-5.3-Flash
Zhipu AI
65.1%
— Report Unverified

What now?

You've seen how models score on Workspace-Bench-Lite — here's what to do next.