Private by default
Keep your work on your machine. Local models let prompts, responses, and files be processed on-device instead of being sent to a remote provider.
Local AI for private work.
Run supported GGUF models with your files in a native workspace. Chats, attachments, downloaded local models, and settings stay on your device.
- Sensitive work
- No cloud billing
- Work offline
Run the local model families your work calls for.
Aitopus supports compatible GGUF models from these families, so private requests can stay on-device when a model is installed.
Qwen
LFM
GPT-OSS
Gemma
Granite
Phi
GLM
Why local models
Works offline
Once downloaded, local models are available without an internet connection. Use AI while traveling or working from restricted networks.
No per-token cost
Run local models on your own hardware without per-token charges, API keys, or cloud billing for local chats.
Device-level control
Choose which models run on your device and keep your conversations on your machine.
Manage models locally
Install, use, and delete local models from your device. Stay in control of storage, model choice, and your local setup.
How it works
01
Pick a local model
Browse compatible local models in Aitopus and compare each model's strengths, size, and fit for your device.
Qwen 3.6 27B
RecommendedPowerfulFlagship reasoning model for coding and multilingual chat.
256K context2 quantizations1 downloadingFits
Gemma 4 31B IT
RecommendedPowerfulDense Google model with vision and broad language support.
256K context2 quantizationsFits
Gemma 4 E2B IT
FastLightweight model for quick local chat and reasoning.
128K context4 quantizations1 installedFits
GLM 4.7 Flash
PowerfulFastMoE model from Z.ai for reasoning, coding, and bilingual chat.
203K context2 quantizationsFits
02
Choose the right download for your device
Many local models are offered in multiple quantizations. Smaller quantizations use less memory and storage, and usually run faster. Larger ones may preserve more quality, but need more capable hardware. Aitopus helps you pick the version that fits your device.
OpenAI GPT-OSS 20B (F16)
13.8 GBFitsF16Max 128K context
OpenAI GPT-OSS 20B (Q4_K_M)
Fast11.6 GBFitsQ4_K_MMax 128K context
OpenAI GPT-OSS 20B (Q8_0)
Fast12.1 GBFitsQ8_0Max 128K context
03
Select it from the model picker
Once the download finishes, the model appears in Aitopus alongside your other available models. Select it from the model picker to start chatting.
Qwen 3.6 35B A3B (UD-Q4_K_M)
Apple Intelligence
Apple
Granite 4.0 H 1B
Q4_K_M
Qwen 3.6 27B
Q8_0
Gemma 4 E4B IT
BF16
Qwen 3.6 35B A3B
UD-Q4_K_M
04
Start chatting locally
Use the Aitopus chat interface with a model running on your own machine.
On device / llama.cpp
What can I help with?
This chat is routed to the local model selected above.
Message Aitopus
05
Manage storage anytime
See which local models are installed, how much space they use, and remove them whenever you need the space back.
Qwen 3.6 27B (Q8_0)
28.6 GBTight fitQ8_0Max 256K context
Context length
128KChoose how much conversation history the model can use.
4K256K
Qwen 3.6 27B (UD-IQ2_M)
10.8 GBFitsUD-IQ2_MMax 256K context
Context length
32KChoose how much conversation history the model can use.
4K256K