Own your own intelligence.
Enterprise AI in Australian jurisdiction, never trained on your data. Open Source AI LLM models served in Australia. Use them through our Inference API, rent them in the PrivChat workspace, or own the on-premise hardware they run on. No prompt retention, AUD billing.
Hardware Appliances
- AIC Compact (AIC-Compact-100): Highest model capacity per dollar. For organisations that want the largest models running quietly on a desk. Runs Gemma 4 26B MoE, Gemma 4 31B IT, GPT-OSS 120B (MoE), Ornith 1.5 35B, Qwen 3.8 Flash Next 125B. 128GB unified memory, 256 GB/s, 38 dBA at 1m. $6,500 - $8,500 AUD ex-GST.
- AIC Studio (AIC-Studio-200): Built for documents. Roughly twice the memory bandwidth of the Compact, with a native vision-language model. Runs Qwen 3.8 27B, Gemma 4 26B MoE. 32GB or 64GB GDDR6 (single or dual GPU), 608 GB/s. $7,900 - $12,500 AUD ex-GST.
- AIC Duo (AIC-Duo-600): Runs everything we host, including DeepSeek V4 Flash, inside your own rack. Runs DeepSeek V4 Flash 0731, plus every smaller model in our catalog. 192GB VRAM across dual RTX PRO 6000, 1.79 TB/s. Price on application.
Supported Open-Weights Models
| Model | What it's for |
|---|---|
| GLM-5.3-Flash (Vision) (320B MoE, 18B active) | Visual coding, documents, and agent work at low cost |
| Qwen 3.8 Flash Next (125B MoE, 6B active) | Fast coding, office, and long-context agent tasks |
| DeepSeek V4 Flash 0731 (284B MoE, 13B active) | Frontier-class reasoning and coding |
| Qwen 3.8 27B (vision-language) | Documents, scanned forms, charts |
| Gemma 4 31B IT | General chat, structured output |
| Gemma 4 26B MoE | Fast, low-latency workloads |
GPU Instances & Hosting Availability
Dedicated GPU instances across PRO 6000 (96GB), RTX 5090 and RTX 4090 accelerators. We offer managed hosting with exclusive utilisation (no root access required) and full bare-metal root access under long-term commitments.
Contact
Melbourne Workshop & Office: Level 1, 41-43 Stewart Street, Richmond VIC 3121, Australia. Phone: +61 451 417 496.