New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
GMKtec EVO-X3
Runs downloadable LLMs up to ~195B at Q4, ~37.1-45.1 tok/s on an 8B model.
What you can run
At 8k context, FP16 KV. Full-speed tok/s holds up to "max ctx"; beyond it the context spills to SSD swap and drops to the degraded figure.
| Model size | FP16 | Q8 | Q4 |
|---|---|---|---|
| 3B |
27.7-33.7 t/s ≤1643k ctx · then ~1.1-1.3 (slow) |
55.5-67.4 t/s ≤1689k ctx · then ~2.2-2.6 (slow) |
99-120.4 t/s ≤1709k ctx · then ~3.9-4.7 (slow) |
| 7-8B |
10.4-12.6 t/s ≤745k ctx · then ~0.4-0.5 (slow) |
20.8-25.3 t/s ≤806k ctx · then ~0.8-1 (slow) |
37.1-45.1 t/s ≤833k ctx · then ~1.5-1.8 (slow) |
| 13B |
6.4-7.8 t/s ≤535k ctx · then ~0.3-0.3 (slow) |
12.8-15.6 t/s ≤615k ctx · then ~0.5-0.6 (slow) |
22.9-27.8 t/s ≤650k ctx · then ~0.9-1.1 (slow) |
| 32B |
2.6-3.2 t/s ≤190k ctx · then ~0.1-0.1 (slow) |
5.2-6.3 t/s ≤312k ctx · then ~0.2-0.2 (slow) |
9.3-11.3 t/s ≤365k ctx · then ~0.4-0.4 (slow) |
| 70B | won't run |
2.4-2.9 t/s ≤133k ctx · then ~0.1-0.1 (slow) |
4.2-5.2 t/s ≤227k ctx · then ~0.2-0.2 (slow) |
| 120B | won't run | won't run |
2.5-3 t/s ≤142k ctx · then ~0.1-0.1 (slow) |
| 400B | won't run | won't run | won't run |
Estimate: tok/s ≈ bandwidth ÷ (params × bytes/param) × efficiency (GMKtec). Cyan = fits ~115.2 GB usable. Beyond "ctx", KV spills to SSD swap (rough degraded speed shown).
Run popular models on the GMKtec EVO-X3
About the GMKtec EVO-X3
GMKtec EVO-X3 is a unified-memory system from GMKtec, 2026. For local genAI the numbers that matter are its 128 GB (how big a model + context fits) and 256 GB/s bandwidth (generation speed).
Frequently asked questions
What size LLM can the GMKtec EVO-X3 run?â–¶
With ~115.2 GB usable it runs up to ~195B at Q4 (~55.5B at FP16) at 8k context. Bigger spills into SSD swap and slows sharply.
How many tokens/sec does the GMKtec EVO-X3 generate?â–¶
Generation is bandwidth-bound. At 256 GB/s an 8B model at Q4 runs ~37.1-45.1 tok/s (estimate).
Why does long context slow it down?â–¶
The KV cache for the context must also fit fast memory; once weights+KV exceed 115.2 GB it spills to SSD swap and throughput collapses. Each row shows the max usable context.