All tasks / Agents & work / Workflow automation
Best local AI for workflow automation
Open-weights models and tools that run entirely on your own hardware, ranked from public benchmarks. Local tools are rated by the best open model they run.
BTop ranked: DeepSeek-V4.1-Flash-1017 (DeepSeek), 66.6% on AutomationBench-AA (Artificial Analysis).12 options listed · data verified 19 Oct 2026
Ranked list
| Tier | Name | Approach | Runs | Rated by | Price | Skill |
|---|---|---|---|---|---|---|
| B | DeepSeek | Models | Local | 66.6% on AutomationBench-AA (Artificial Analysis) | ~320 GB VRAM | — |
| B | Z.ai | Models | Local | 62.2% on AutomationBench-AA (Artificial Analysis) | ~425 GB VRAM | — |
| B | Z.ai | Models | Local | 60.4% on AutomationBench-AA (Artificial Analysis) | ~185 GB VRAM | — |
| C | Xiaomi | Models | Local | 58.6% on AutomationBench-AA (Artificial Analysis) | ~585 GB VRAM | — |
| C | Moonshot AI | Models | Local | 58.3% on AutomationBench-AA (Artificial Analysis) | ~1600 GB VRAM | — |
| C | StepFun | Models | Local | 51% on AutomationBench-AA (Artificial Analysis) | ~346 GB VRAM | — |
| C | Meta | Models | Local | 49.7% on AutomationBench-AA (Artificial Analysis) | ~470 GB VRAM | — |
| C | Reflection AI | Models | Local | 48.5% on AutomationBench-AA (Artificial Analysis) | ~287 GB VRAM | — |
| C | Alibaba (Qwen) | Models | Local | 48.2% on AutomationBench-AA (Artificial Analysis) | ~17 GB VRAM | — |
| C | Thinking Machines Lab | Models | Local | 5% on AutomationBench-AA (Artificial Analysis) | ~557 GB VRAM | — |
| C | MBZUAI Institute of Foundation Models | Models | Local | 37.2% on AutomationBench-AA (Artificial Analysis) | ~218 GB VRAM | — |
| C | NVIDIA | Models | Local | 14.8% on AutomationBench-AA (Artificial Analysis) | ~72 GB VRAM | — |
Tiers come only from public benchmarks (AutomationBench-AA (Artificial Analysis)); tools without a result of their own inherit the tier of the best model they offer. How we rank.