New updates and improvements
V7 Go

You can now filter your data natively in Go, which was one of our top feature requests. Click the filter icon in the lower left corner to add filters and combine them using all/any/none logic for Text or Select properties.

For classification tasks in Select properties, the AI now provides explanations when it can't find a matching option. This improvement enhances transparency in AI workflows and helps you spot potential data taxonomy issues at a glance, as well as creating a feedback loop to iterate on your prompts.

Soon, we’ll add more granularity so you can define actions when the AI can’t find a match and show reasoning for successful matches too.

We've solved the "content too long" error for documents exceeding model input token limits. Our new chunking mechanism allows V7 Go to process extensive, multi hundred page documents like SEC Form S-1 filings, lengthy technical documentation, or court hearing transcripts.

This alpha feature is available behind a feature flag. Contact us with your use case, and we can enable it for your V7 Go account.

We have some great news! Our V7 Go Summer Update is now live, packed with exciting new features to level-up your document processing, data extraction, and AI automation workflows.

Watch the video below to learn how to enable trustworthy AI in your company and automate even the most challenging processes:

Explore our latest features:

• Ask Go. Set up projects in seconds using natural language

• AI Citations. Verify AI outputs at a glance with PDF highlights

• Collections. Organize unstructured data into custom tables

• Web Properties. Extract data from URLs and perform web searches

• Graphs (Beta). Interpret charts and convert them to tables with multimodal AI

• Audio Analysis. Process audio files in batches for summaries and transcriptions

• Supervised Fine-Tuning & RLHF. Teach AI to solve specialized tasks

Plus, we've added:

• New AI models: GPT-4o Mini, Claude 3.5 Sonnet, and more

• Python Code Tool for custom operations using NumPy, Pandas, and more

• Improved CSV import with granular column and row mapping

• Bring your own API key option for enterprise users

All of these features are designed to make your AI workflows more efficient, controllable, and secure.

The new GPT-4o mini model from OpenAI is now available to all V7 Go users. This compact version offers performance similar to the standard GPT-4o, but at a fraction of the cost.

• 60% cheaper than GPT-3.5 Turbo and 96% cheaper than GPT-4o.

• 128K token context window with up to 16K output tokens per request.

• Handles both text and vision inputs, with audio support coming soon.

GPT-4o mini is ideal for automation tasks that require multiple model calls and for processing large volumes of documents.

Struggling with unreliable AI outputs? Our new V7 Go feature ensures your AI responses are accurate by linking them to specific citations in your source files, such as PDF documents.

Visual grounding shows exactly where your AI found its information. This allows for instant verification of AI-generated responses, removes errors, and reduces hallucinations. Highlighted AI citations make your document processing workflows more transparent and trustworthy. They also slash QA time from hours to minutes.

To use the feature:

1. Toggle Grounding in your AI property setup

2. Inspect a selected entity and click the lightbulb icon

3. Click on sentences to navigate to the relevant citation

The feature is now available to all V7 Go users. Find out more about visual grounding in our documentation.

Based on valuable feedback from our early users like you, we've decided to double the number of Tokens available in the free plan. Now, new users receive 2x as many Go Tokens to power their tasks in V7 Go.

If you already have an existing account and want to redeem your extra Tokens, simply contact us, and our team will assist you.

To fully unlock the potential of Go for automating tasks at scale, you can also switch to one of our pro plans.

You can now execute Python code as part of your V7 Go workflows. Instead of using AI models for all tasks, add the Python Tool and write custom code snippets.

• Access Python libraries. You can use the Python Tool for specialized tasks, such as parsing web-scraped content with BeautifulSoup4 or performing data manipulation with Pandas.

• Get predictable and deterministic results. Generate consistent and controllable outputs to make your workflows more reliable in a production environment.

• Reduce token consumption. Traditional code can sometimes provide a more efficient solution and will help you reduce the unnecessary expense of AI models.

How to use the Python Tool

1. Add a new property. Choose the correct type. The Python Tool supports Text, JSON, Single Select, and Multi Select properties.

2. Select Python Tool. From the list of tools, select Python Tool.

3. Add variables. Use @ to select and add variables based on other properties.

For a detailed guide, visit our documentation.

The new Library lets you add documents and images as guidelines for your AI models.

Use these files as additional inputs to improve the performance of LLMs when specialist, context-dependent knowledge is required.

• Add your company-specific context and knowledge to foundation models

• Use images, photos, or scans as reference for quality control or defect detection

• Cross-reference your input data with custom policies, guidelines, or dictionaries

How to use the Library

Drag and drop PDFs, CSVs, or images into the Library section of the UI. These files will become selectable as inputs for all properties in your V7 Go projects.

Additionally, you can combine these files with the new Python Code Tool to use them as variables.

For more details, visit our documentation.

You can now use the latest AI model from Anthropic in V7 Go for document processing, visual captioning, and much more.

Claude 3.5 Sonnet

• Beats GPT4o on many reasoning benchmarks

• 2x faster than Claude 3.5 Opus

• 1/5th the cost of Opus, and cheaper than GPT4o

To set up a V7 Go project powered by Claude 3.5 Sonnet, edit an AI property and select the model from the Tool list.

If you want to find out more about the model and its performance, visit the official release post. You can also set up your own benchmarking tests directly in V7 Go.

One of our top feature requests post-launch is now live in V7 Go.

Go has always supported CSV upload into a single field, which is powerful if you want an AI model to reason across an entire CSV at once.

But lots of users have asked to import a CSV which then unfurls into a project, where AI models can reason over individual columns in the same structure as the CSV.

We're excited to announce that V7 has upgraded to GPT-4o, the latest advancement in AI technology. Today, GPT-4o is much better than any existing model at understanding and discussing the images you upload. For example, you can now take a picture of a menu in a different language and talk to GPT-4o to translate it.

What's new

• Enhanced Capabilities and Accessibility: GPT-4o offers improved language capabilities in quality and speed, supporting over 50 languages for sign-up, login, user settings, and more. It is designed to make advanced AI tools more accessible and useful to a global audience.

• Expanded Features for Free Users: GPT-4o provides ChatGPT Free users with access to advanced features such as GPT-4 level intelligence, web-based responses, data analysis with chart creation, photo discussions, file uploads for summarizing, writing, or analyzing, and the ability to discover and use GPTs and the GPT Store. These features significantly enhance the user experience, even for those not on paid plans.

• Improved User Experience with Desktop App: The new ChatGPT desktop app for macOS integrates seamlessly into the user's workflow, offering instant access to ChatGPT via a simple keyboard shortcut. It supports voice conversations and features like screenshot discussions, making it a powerful tool for brainstorming, preparing for interviews, and more. The app will also be available for Windows later this year.

We continue to be committed to leveraging cutting-edge technology to provide you with the best possible experience and insights. Enjoy the enhanced capabilities of Go powered by GPT-4o!