Can Surface Laptop Ultra · RTX Spark N1X (18-core) · 24 GB run Llama 3.3 70B Instruct?
24 GB unified · 300 GB/s memory bandwidth · Llama 3.3 70B Instruct has 70.6B parameters.
Short answer
No. It doesn't fit in memory.
Needs ≈37 GB but only ≈16 GB is available after the OS, other apps and Open WebUI.
Why it doesn’t fit
Even its smallest practical version (Q3_K_M) needs about 37 GB for the model, its context memory and the runtime. After Windows (≈4.5 GB), a browser with a few tabs (≈3.0 GB) and Open WebUI (≈1.0 GB), Surface Laptop Ultra · RTX Spark N1X (18-core) · 24 GB has about 16 GB left, so it’s 21 GB short.
Ways to make it work
Close everything else
Not enough on its own: with every other app closed it’s still about 18 GB short.
Pick a smaller model from the same family
Get more memory
The same machine with 128 GB handles it: Acceptable.
Check your exact setup
These pages assume typical settings. A shorter context or a lighter chat app can change the answer.
Check Surface Laptop Ultra · RTX Spark N1X (18-core) · 24 GB
The same applies to every workload we rate (chat, coding questions, coding in a repository, agentic coding and long documents): they all need at least as much memory as chat.
Other models for Surface Laptop Ultra · RTX Spark N1X (18-core) · 24 GB
The best-rated models for chat on this machine.
- gpt-oss-20b50–85 tok/sExcellent
- MiniCPM5 2B55–90 tok/sExcellent
- Gemma 4 26B-A4B60–95 tok/sExcellent
- Qwen3.5 4B35–55 tok/sExcellent
- Qwen3.5 9B18–30 tok/sExcellent
- Gemma 4 12B20–35 tok/sExcellent
Cheapest machines for Llama 3.3 70B Instruct
Hardware rated comfortable or better for chat, lowest price first.
- Desktop PC · 2× GeForce RTX 3090 (48 GB) · 128 GB RAM$2,884 · 11–18 tok/sComfortable
- Mac Studio M5 Ultra 128 GB$5,499 · 10–16 tok/sComfortable
- Mac Studio M5 Ultra 256 GB$6,299 · 10–16 tok/sComfortable
- Mac Studio M5 Ultra 512 GB$7,099 · 10–16 tok/sComfortable
- Desktop PC · RTX PRO 6000 Blackwell · 128 GB RAM$17,284 · 17–25 tok/sExcellent
Same chip, other memory: 128 GB
Different tool, context size or workload?
These pages use typical settings. The full check models your exact setup.