Builds / None (CPU only)
Intel NUC8i7BEH (Core i7-8559U, CPU only): Nemotron 3.5 Bolt E4B Q4_K_M at 5.0 tok/s, Qwen3.5 4B 5.63, Qwen3.5 9B 3.07 (llama.cpp)
Last updated:
Source: mustafa.net ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
- GPU(s)
- None (CPU only)
- VRAM / unified memory (GB)
- 0
- CPU
- Intel Core i7-8559U (NUC8i7BEH, 2018)
Amazon · eBay (used) · Newegg - Motherboard
- Intel NUC8i7BEH
Amazon · eBay (used) · Newegg - RAM
- DDR4
Amazon · eBay (used) · Newegg - PSU
- not stated
- Other
- CPU only, llama.cpp b11561 (official Docker image)
- Reported total cost
- not stated
- Power
- not stated
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Qwen3.5 9B | Q4_K_M | llama.cpp b11561 server (CPU) | 3.07 | not stated | context: 8,192 ctx📏 Same test, 3 x 256 tokens at temperature 0, median |
| Nemotron 3.5 Bolt E4B legacy | Q4_K_M (NVIDIA official GGUF, 6.18 GB) | llama.cpp b11561 server (CPU) | 5.0 | not stated | context: 8,192 ctx📏 llama-server, 32-token warm-up, then 3 x 256 tokens at temperature 0, median (5.03-5.04); prompt processing 30 tok/s; measured 7 Nov 2026 |
| Qwen3.5 4B legacy | Q4_K_M | llama.cpp b11561 server (CPU) | 5.63 | not stated | context: 8,192 ctx📏 Same test, 3 x 256 tokens at temperature 0, median |
Mustafa.net homelab test (published 12 Nov 2026). Same test and llama.cpp builds as the author's RTX 2060 Super and Mac mini M4 builds on RigBase. Nemotron 3.5 Bolt E4B is NVIDIA's 8B-total / 4B-effective model post-trained from Gemma 4 E4B; it ran 64% faster than Qwen3.5 9B here because its two large embedding tables are lookups.
Source: https://mustafa.net/2026/11/12/nemotron-3-5-bolt-run-locally/
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)