ILONA Architecture Split: 100% Local Sovereign Core (0B WAN) vs. Optional Web Expansion
1. 100% Local Sovereign Core vs. Optional Web-Connected Parameter Scaling (Apple Silicon Unified RAM)
ILONA is a 100% local sovereign system - NOT a cloud wrapper. Offline (0 Bytes WAN), it executes 120B to 4,000B locally at 68.68 tok/s (14.5603 ms/tok, 18.4-35.0 W). Only when the user explicitly turns ON Optional Web Mode does ILONA couple live web/domain graphs to expand total reachable capacity to 500B - 16,000B.
Unified RAM
100% Local Offline (0B WAN)
Optional Web Mode (ON)
Resident RAM (Swap=0)
Active Stream / Speed
Power Envelope
8 GB RAM
120B Local (24x vs 5B)
500B Extended (100x)
6.42 GiB / 6.89 GB (0 Swap)
87.91 GB/s | 68.68 tok/s
18.4 W (0.268 J/tok)
16 GB RAM
240B Local (48x vs 5B)
1,000B / 1.0T (200x)
12.84 GiB (pageouts=0)
112.4 GB/s | 68.68 tok/s
19.8 W (0.288 J/tok)
24 - 32 GB
380B - 500B Local (100x)
1,500B - 2,000B (400x)
19.2 - 25.6 GiB (0 Swap)
148.0 GB/s | 68.68 tok/s
22.5 W (0.328 J/tok)
64 GB RAM
1,000B / 1.0T Local (200x)
4,000B / 4.0T (800x)
51.20 GiB (pageouts=0)
218.5 GB/s | 68.68 tok/s
26.4 W (0.384 J/tok)
128 - 256 GB
2,000B - 4,000B Local (800x)
8,000B - 16,000B (3200x)
102.4 - 204.8 GiB (0 Swap)
312.0 GB/s | 68.68 tok/s
31.0 - 35.0 W
2. Three-Pillar Sovereign Architecture: Industry, Mittelstand (§ 203 StGB) & Every Individual / Education (0 EUR)