Recommended Local AI Models for RTX 5060 Ti (VRAM Presets)
The RTX 5060 Ti class is well suited to local AI agents: capable enough for useful workflows, but easy to overload with oversized models. Below are practical presets for pairing Ollama with desktop assistants such as Open Friday.
Treat model tags as starting points. Quantization, context length, and concurrent applications affect what fits in VRAM.
Preset A — Balanced daily use
Default for chat and light agent work:
| Role | Model tag | Notes |
|---|---|---|
| Chat / agent | qwen3.5:9b | Strong default quality |
| Fast router | qwen3.5:4b | Faster for classification and light tasks |
| Embeddings | nomic-embed-text | Memory / RAG search |
| Vision | qwen3-vl:8b-instruct | Screen-aware steps |
Goal: remain responsive while leaving headroom for browser, IDE, and the desktop agent.
Preset B — Latency first
When response time matters more than peak reasoning quality:
- Chat:
qwen3.5:4b - Vision: a smaller VL quant if available (for example, 7B-class Q4)
- Keep embeddings loaded; unload unused chat models between sessions
Preset C — Higher quality with VRAM headroom
When the machine is not heavily multitasking:
- Optional coding:
qwen3-coder:30bonly if VRAM allows - Otherwise prefer 7B–9B chat with 7B–8B vision — more reliable when agents need concurrent tools
Guidelines for 5060 Ti–class cards
- One heavy model at a time — chat or a large coder model, not both plus a vision model
- Reserve capacity for vision — screenshots and UI automation increase memory use
- Measure under real load — monitor VRAM during an actual Friday computer-use session
- Prefer documented quantizations from Friday’s docs unless you have a specific reason to diverge
Suggested pull list
ollama pull qwen3.5:9b
ollama pull qwen3.5:4b
ollama pull qwen3-vl:8b-instruct
ollama pull nomic-embed-text
Then validate with friday doctor if you use Open Friday.
Why businesses choose Devoids
Devoids typically delivers 10–30% lower cost than typical market rates for comparable custom software, SaaS, and mobile work — without cutting corners. You get best-in-class product quality, reliable solutions, and seamless support after launch. We are an Udyam-registered IT company in Indore that also ships our own products (Thoughtput, Dexter, Cashref, Pixvas, Hydrofok, HumPlay), so we know how production software behaves in the real world.