Local AI.
Fully Offline.
The FlowSIM Pro assistant runs entirely on your computer. No cloud, no account. Your projects never leave your machine.
About the AI Assistant
FlowSIM Pro ships with a capable default model (Qwen3.5 4B) that is installed automatically. You can use the assistant right away, no download required. If your computer has enough memory, the larger models on this page give noticeably better answers. After the one-time download, everything runs fully offline.
How the Assistant Works
The assistant runs as a separate local process inside FlowSIM Pro, powered by the open llama.cpp engine. Your prompts and your project data stay on your computer. After the one-time model download, no internet connection is required.
You can chat with it in English or German. It answers questions about FlowSIM Pro and your scene, and it executes commands you type in plain language, for example:
- “Create a box of 2 x 3 x 2.5 meters” adds the primitive to the scene.
- “Set up a wind comfort study, 10 m/s from southwest” configures the Wind Wizard with these parameters.
- “Set the simulation end time to 400” changes the OpenFOAM solver settings.
- “Delete all radiators” matches objects by name and asks for confirmation first.
The language model itself never edits your files or your scene directly. It only translates your request into parameters for the same functions that the regular user interface uses. Every action therefore follows the same validation, confirmation, and undo rules as a manual click.
Scene commands require the full memory profile, which FlowSIM Pro enables on machines with more than 16 GB RAM. On smaller machines the assistant still answers questions in a memory-saving chat mode.
How to Install
- Download all parts of the model of your choice (links below).
- Save all parts in the same folder, for example your
Downloadsfolder. - In FlowSIM Pro, open the AI assistant, choose Select local AI model, and pick part 00001 (the file ending in
-00001-of-….gguf). The remaining parts are found automatically.
Qwen3.5 9B
| File | Size | SHA-256 |
|---|---|---|
| Qwen3.5-9B-Q4_K_M-00001-of-00002.gguf | 3.62 GB | 4594d4a74c7cf5665b74981d7478e0515ea3a3b3dd9c9ca8def66a5ddc5c5838 |
| Qwen3.5-9B-Q4_K_M-00002-of-00002.gguf | 1.67 GB | 4563a12e69089ce22467bf0ebb280746fabfcabebc1aea1a026c54117312a739 |
Qwen3.5 27B
| File | Size | SHA-256 |
|---|---|---|
| Qwen3.5-27B-Q4_K_M-00001-of-00005.gguf | 3.62 GB | 62e967d71c2558f44a4f2819222ea7047d4c32da908eb4591707ed2a8511ff11 |
| Qwen3.5-27B-Q4_K_M-00002-of-00005.gguf | 3.61 GB | 6716a2f15be235347373df71c8bf803d1142a26b4ca3037e53c94508f0d5a66b |
| Qwen3.5-27B-Q4_K_M-00003-of-00005.gguf | 3.63 GB | 1af8614cb1062f6c54ab38f7ddfa7f53e47c57b3c184fa072f310a05ff493d93 |
| Qwen3.5-27B-Q4_K_M-00004-of-00005.gguf | 3.62 GB | 35419790a35f06be92a224bd0ef07f5736ee155c60cd841fbe29e8247f8f4f52 |
| Qwen3.5-27B-Q4_K_M-00005-of-00005.gguf | 1.11 GB | 4d1f2042790a27ab440ca131b909b3379ff6f8f5705446b70dc3c8c40cacbff4 |
Verify Your Download
An incomplete download shows the error “Could not load GGUF model” in FlowSIM Pro. In that case compare the checksum of your file with the table above and re-download the affected part. All checksums are also available as SHA256SUMS.txt.
Windows (PowerShell):
Get-FileHash .\Qwen3.5-9B-Q4_K_M-00001-of-00002.gguf -Algorithm SHA256
Linux / macOS:
sha256sum -c SHA256SUMS.txt
License & Attribution
The models are developed by the Qwen team (Alibaba Cloud), quantized to GGUF by Unsloth, and are redistributed here under the Apache License 2.0. A copy of the license is included as LICENSE.txt. Original sources: Qwen3.5-9B-GGUF, Qwen3.5-27B-GGUF.