onpremisai.be Book a first call

Integration service

We install the AI server where your teams already work.

The server becomes a private AI layer connected to the tools already used by developers, business teams and operations.

Developer workspace

Directly inside IDEs

Connect assistants for coding, documentation, refactoring and technical analysis inside common development environments.

Local access

Private chatbot on your site

A local chat interface can be exposed directly through the intranet, with access rules and internal knowledge retrieval.

Enterprise knowledge

Connected to SharePoint and document sources

Index and search controlled company knowledge: SharePoint sites, OneDrive folders, Teams files, Confluence, Nextcloud, NAS shares or S3-compatible storage.

Autonomous workflows

Connected to dedicated agents

Run specialized agents for targeted tasks: code work, document analysis, internal support, ticket preparation and workflow automation.

OpenClaw OpenClaw Hermes Hermes Codex Codex

Model details

Models tracked for deployments.

Each model is evaluated by modality, memory footprint, status and fit with local infrastructure constraints. Depending on your needs, we can recommend the most suitable models.

View all models
The 3 latest validated models
HF Trending

GLM-5.2

Z AI · 2026-06-17

Modality
text → text
Parameters
753B total · active not specified
Size
Not specified
HF Trending

Nemotron 3.5 ASR Streaming

Nvidia · 2026-06-16

Modality
audio → text
Parameters
0.6B total · active not specified
Size
Not specified
HF Trending

MiniMax-M3

MiniMax AI · 2026-06-15

Modality
image, text → text
Parameters
427B total · active not specified
Size
Not specified
The 3 best-known models
Stable

Gemma 4

Google · 2026-04-02

Modality
text, image, video → text, tool, code
Parameters
31B total · 31B active
Size
32gb
Stable

DeepSeek-V4-Pro

DeepSeek AI · 2026-06-07

Modality
text → text
Parameters
862B total · active not specified
Size
Not specified
Stable

Qwen 3.6 VL

Alibaba Cloud · 2026-04-24

Modality
image, text → text
Parameters
35B total · 3B active
Size
24gb