AISSIST Documentation

🚀 Version 12.0 ​

With version 12.0, AISSIST extends the free text chat with flexible file sources – including toggleable documents – and introduces powerful new language models such as "Mistral Large 3", "Gemini 3 Pro", "Gemini 2.5 Flash" and "GPT Image 2.0".

📎 Free Text Chat with Files – Add, Toggle & Reference Sources Directly ​

In free text mode, you can now add files as sources and collaborate with AISSIST even more precisely.

New Possibilities ​

  • Upload documents and media and use them as context for your chat
  • Flexibly toggle sources on and off, depending on which documents are currently relevant
  • Reference files directly in the chat (e.g. "Summarise the PDF" or "Analyse the table in file 2")
  • Once a source is deactivated, the model no longer has access to its content – you retain full control over which information is used in each chat

Example from the "Free Text Chat" Workflow ​

File sources in free text chat – adding and toggling sources

The free text chat with uploaded file sources that can be flexibly activated or deactivated.

🧠 New Models – "Mistral Large 3", "Gemini 3 Pro", "Gemini 2.5 Flash" and "GPT Image 2.0" ​

With version 12.0, additional powerful models are available that you can select depending on your requirements.

Highlights ​

  • Mistral Large 3 – strong for complex texts, analyses and longer content
  • Gemini 3 Pro – versatile model for demanding knowledge and creative tasks
  • Gemini 2.5 Flash – optimised for high speed with solid response quality
  • GPT Image 2.0 (beta) – OpenAI's current image generation model with improved prompt adherence and image quality; high image fidelity with clean composition, lighting and fine details

You can use these models flexibly in your use cases to find the right mix of quality, speed and contextual understanding for each task.

💡 Quality of Life – Improvements in Detail ​

In addition to the new main features, various improvements have been implemented:

  • Documents in templates can now be stored for every model – no longer only for GPT models
  • Colour picker in image generation is now also available for additional models
  • Web search has been enabled for all models
  • Reasoning selection is now available across all models – depending on the use case, you can select how much the model should "think"
  • Transcription now supports files up to 512 MB for video and audio content

Example: Reasoning Setting ​

Cross-model reasoning selection in AISSIST

The reasoning selection is now available for all supported models.

For further details on all new features, see the full documentation.