What hardware do I need?
Hardware that delivers the experience you want — not just loads the model
Pick the model and describe the workload. We rate every machine in the database for that exact workload and group them by whether they meet your target.
1 · Model
2 · Use case
Agentic coding
Chat & writing
Coding
Documents & research
Multimodal
Automation & serving
3 · Tool & runtime
Tool: OpenCode
4 · Workload
Project size, agent behaviour, other apps
5 · Desired experience
6 · Budget & vendors (optional)
Prices are approximate launch prices for the whole system.
gpt-oss-20b · Agentic coding / medium repository
Target: comfortable · 83 machines meet it
Meets your comfortable target (83)
Cheapest first.
Desktop PC · GeForce RTX 5060 Ti 16GB · 32 GB RAM
16 GB VRAM (448 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · Ollama
- Generation
- ≈85 tok/s
- Per step
- ≈8 s
- Headroom
- 2.1 GB
Desktop PC · GeForce RTX 4060 Ti 16GB · 32 GB RAM
16 GB VRAM (288 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · Ollama
- Generation
- 40–70 tok/s
- Per step
- ≈10 s
- Headroom
- 2.1 GB
Desktop PC · Radeon RX 9070 XT · 32 GB RAM
16 GB VRAM (640 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · Lemonade
- Generation
- 130–170 tok/s
- Per step
- ≈5 s
- Headroom
- 2.1 GB
Desktop PC · Intel Arc Pro B60 24GB · 32 GB RAM
24 GB VRAM (456 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 50–85 tok/s
- Per step
- ≈15 s
- Headroom
- 10.0 GB
Desktop PC · Radeon RX 7900 XT · 32 GB RAM
20 GB VRAM (800 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · Lemonade
- Generation
- 120–200 tok/s
- Per step
- ≈5 s
- Headroom
- 6.1 GB
Ryzen AI Max+ 395 mini PC / Framework Desktop · 64 GB
64 GB unified memory · 256 GB/s
- Quant / runtime
- MXFP4 · Lemonade
- Generation
- 55–75 tok/s
- Per step
- ≈15 s
- Headroom
- 37 GB
Desktop PC · Intel Arc Pro B60 24GB · 64 GB RAM
24 GB VRAM (456 GB/s) + 64 GB RAM
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 50–85 tok/s
- Per step
- ≈15 s
- Headroom
- 10.0 GB
Desktop PC · GeForce RTX 5070 Ti · 32 GB RAM
16 GB VRAM (896 GB/s) + 32 GB RAM
- Quant / runtime
- MXFP4 · Ollama
- Generation
- ≈140 tok/s
- Per step
- ≈5 s
- Headroom
- 2.1 GB
Mac mini M2 Pro 32 GB
32 GB unified memory · 200 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 35–50 tok/s
- Per step
- ≈20 s
- Headroom
- 8.9 GB
Meets the acceptable target (8)
Usable, with noticeable compromises for this workload.
MacBook Air M4 32 GB
32 GB unified memory · 120 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 20–30 tok/s
- Per step
- ≈30 s
- Headroom
- 8.9 GB
MacBook Air M5 32 GB
32 GB unified memory · 153 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 20–35 tok/s
- Per step
- ≈20 s
- Headroom
- 8.9 GB
MacBook Pro M1 Pro 32 GB
32 GB unified memory · 200 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 35–50 tok/s
- Per step
- ≈20 s
- Headroom
- 8.9 GB
MacBook Pro M4 32 GB
32 GB unified memory · 120 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 20–30 tok/s
- Per step
- ≈30 s
- Headroom
- 8.9 GB
MacBook Pro M5 32 GB
32 GB unified memory · 153 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 20–35 tok/s
- Per step
- ≈20 s
- Headroom
- 8.9 GB
Mac mini M4 32 GB
32 GB unified memory · 120 GB/s
- Quant / runtime
- MXFP4 · LM Studio
- Generation
- 20–30 tok/s
- Per step
- ≈30 s
- Headroom
- 8.9 GB
Laptop · Ryzen AI 9 HX 370 (Radeon 890M) · 32 GB
32 GB unified memory · 120 GB/s
- Quant / runtime
- MXFP4 · Lemonade
- Generation
- 18–30 tok/s
- Per step
- ≈35 s
- Headroom
- 8.0 GB
Desktop PC · Ryzen 9 9950X (no GPU) · 64 GB DDR5
64 GB RAM · 89 GB/s · CPU only
- Quant / runtime
- MXFP4 · llama.cpp (llama-server)
- Generation
- 11–18 tok/s
- Per step
- ≈50 s
- Headroom
- 37 GB
Can run, but below your target (19)
The model loads, but the experience falls short for this workload.
Cannot run (18)
Not enough usable memory, or no compatible runtime.