We’re excited to announce the release of Nomic GPT4All v3.4.0, packed with exciting updates including new faster models, expanded filetype support, and several improvements to enhance your experience!
This release introduces the LLaMa 3.2 Instruct 1B and 3B models, offering state-of-the-art performance on lower-end devices. With a context length of 128K tokens, these models excel in tasks such as summarization, instruction following, and rewriting, running efficiently on edge devices.
These 1B and 3B models are comparable in capability to 13B parameter models from April 2023, making them ideal for users seeking powerful performance in smaller, more accessible models.
We've expanded filetype support to include Excel and Word documents. You can now attach Excel spreadsheets directly into your chat sessions and add docx files to your LocalDocs collections. This allows GPT4All to process and contextually understand these files, offering deeper insights and better responses. Learn more about using GPT4All with Excel spreadsheets in the cookbook.
We've also made improvements to the LocalDocs retrieval algorithm, resulting in more accurate references for your queries. This enhancement ensures that GPT4All delivers more relevant and precise information from your local document collections.
We’ve also resolved bugs with the "Redo last chat response" button and issues with continuing chats from previous sessions, ensuring smoother operation.
With the release of Nomic GPT4All v3.4.0, we’re committed to improving your experience with faster models, better file support, and enhanced accuracy. Try out the new LLaMa 3.2 models on your devices today and explore all the latest features!