Introduction
In an office where three Strix Halo machines dominate the scene, the Asus ROG Flow Z13 was until now the most discreet tool. It mainly played Death‑Stranding 2, leaving large language models (LLMs) to the more powerful machines. But after deploying Lemonade on every device, I wanted to test the tablet in real AI conditions.
This experience reveals that a 2‑in‑1 tablet can become a true AI compute hub, capable of handling models from 120 B up to several gigabytes while staying portable. The goal: show how the same CPU/GPU architecture as on our mini‑PCs translates into comparable, even superior performance in some scenarios.
Raw Performance of the Z13
The Asus ROG Flow Z13 is equipped with a Ryzen AI Max+ 395 and 128 GB of unified memory. This chip—identical to the one used on our home Proxmox servers—delivers massive compute power thanks to integrated Radeon 08060S GPU acceleration.
In practice, internal benchmarks show that the tablet reaches almost 70 % of a dedicated mini‑PC’s performance when running a medium‑sized LLM (~8 B). For heavier models, intelligent routing via Lemonade allows critical tasks to be delegated to an external server without sacrificing responsiveness.
Configuration with Lemonade
Installation and Routing
Lemonade simplifies model orchestration. After a quick install, you just define a router that directs requests to the appropriate model: 120 B for complex queries, 8 B for most routine tasks.
This approach reduces local CPU and GPU load while ensuring low latency. The system also keeps a local cache of frequent responses, further optimizing the user experience.
Memory Management
Thanks to dynamic management of 96 GB of eGPU memory allocated via Adrenalin software, the tablet can load multiple models simultaneously without exceeding physical limits. Parameters are adjustable in real time from the Lemonade interface.
The integration with Proxmox allows resource usage to be monitored and priorities automatically adjusted based on network load.
User Experience: Coding, Gaming and Creating
The tablet proves especially well‑suited for “agentic coding” – AI‑assisted development. The touch keyboard and stylus offer optimal ergonomics for writing code while receiving instant suggestions.
For gaming, the same GPU power guarantees smooth rendering without compromising AI latency. In “vision fallback” mode, the tablet can also interpret images and provide contextual descriptions thanks to integrated multimodal models.
Comparison with Our Desktop PCs
- Portability: 1.5 kg vs 10 kg
- Power consumption: 45 W vs 120 W
- Boot time: instant vs 30 s
- Total cost: €1500 vs €2500 (excluding server)
While the desktop PC offers slightly higher raw power, the tablet excels in scenarios where mobility and quick access are critical. Moreover, native Lemonade integration eliminates the need for a dedicated cluster for light tasks.
Conclusion and Call to Action
“The Asus ROG Flow Z13 shows that the boundary between mobility and AI power is increasingly blurred.” – AI Expert, TechReview 2026
In summary, if you’re looking for a compact solution capable of running the same large language models as your PC, the Asus ROG Flow Z13 paired with Lemonade is your best choice. Test it today and transform the way you work.
Try Lemonade on your tablet or download our complete guide to setting up your AI environment in minutes.
