The first time you open 5090Ai, it checks that your PC is ready and helps you choose a model.
Check your system
5090Ai checks Windows, your graphics card, the NVIDIA driver, your memory and whether the model files are on your PC. Green ticks mean you're good to go.
5090Ai
If something needs attention, fix it (for example, update the driver from nvidia.com) and click Check again. Then click Continue.
Choose a model
5090Ai includes two models. Both run entirely on your graphics card and give the original model's exact answers:
- Qwen3.8-27B · Smartest: best for hard reasoning, maths and big coding tasks (~304 tokens per second).
- Qwen3.6-35B-A3B · Fastest: best for everyday chat, long documents and quick drafts (~782 tokens per second).
5090Ai
Not sure? Start with Qwen3.6-35B-A3B. You can switch at any time from the chat screen.
Load it
Click Load Qwen3.6-35B-A3B (or the model you picked). Tick Load automatically at start if you want 5090Ai to load it every time it opens.
Loading takes about 45 seconds. You can follow the progress, or click Cancel to stop.
Start chatting
When the model is ready, the chat screen opens. That's it: you're set up.
5090Ai
