Can Desktop PC · Ryzen 9 9950X (no GPU) · 64 GB DDR5 run Llama 3.3 70B Instruct?
64 GB RAM, no GPU · Llama 3.3 70B Instruct has 70.6B parameters.
Short answer
It loads, but it's too slow even for chat.
Best quantization for chat: Q5_K_M (47 GB of weights) on llama.cpp (llama-server). Generation: 0.7–1.2 tok/s, leaving about 0.8 GB free for your other apps.
How it rates for each workload
Same model and machine, different jobs. Agents and long documents send far bigger prompts than chat, so they need much more speed and memory.
| Workload | Verdict | Quant | Generation | First reply | Context | |
|---|---|---|---|---|---|---|
| Chatwith Open WebUI | Technically Runs | Q5_K_M | 0.7–1.2 tok/s | ≈30 s | 8K | Details |
| Coding questionswith Continue | Technically Runs | Q4_K_M | 0.8–1.4 tok/s | ≈3 min | 16K | Details |
| Coding in a repositorywith Aider | Technically Runs | Q3_K_M | 0.9–1.5 tok/s | ≈13 min | 32K | Details |
| Agentic codingwith OpenCode | Technically Runs | Q3_K_M | 0.9–1.5 tok/s | ≈19 min | 32K | Details |
| Long documents (64K)with Open WebUI | Does Not Fit | — | — | — | — | Details |
Speeds are bandwidth-based estimates; no direct benchmark exists for this pair. “First reply” is the time to read the workload’s opening prompt. How we calculate
Get it running
Copy-paste commands for the Q5_K_M build with a 8K context. Ollama and llama.cpp download the file straight from Hugging Face.
ollama run hf.co/unsloth/Llama-3.3-70B-Instruct-GGUF:Q5_K_M
Downloads 49.95 GB (Q5_K_M). Ollama's default context is small: set OLLAMA_CONTEXT_LENGTH=8192 before `ollama serve` for long prompts.
Other models for Desktop PC · Ryzen 9 9950X (no GPU) · 64 GB DDR5
The best-rated models for chat on this machine.
- LFM2.5 24B-A2B20–35 tok/sExcellent
- MiniCPM5 2B20–35 tok/sExcellent
- Llama 3.2 3B Instruct16–25 tok/sExcellent
- Qwen3.6 35B-A3B11–18 tok/sComfortable
- Qwen3-Next 80B-A3B Instruct15–25 tok/sComfortable
- Gemma 4 26B-A4B10–17 tok/sComfortable
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
Different tool, context size or workload?
These pages use typical settings. The full check models your exact setup.