To remain competitive with leading generative AI companies like Anthropic and OpenAI, Google is enhancing the free tier of its Gemini AI-powered chatbot with new updates. These updates are designed to boost performance and expand availability.

Starting Thursday, Gemini 1.5 Flash—a streamlined multimodal model introduced by Google in May—will be available on the web and mobile platforms in 40 languages across roughly 230 countries. Google asserts that this new model delivers notable improvements in quality and latency, with enhanced capabilities in reasoning and image comprehension.

This update may also lead to cost savings for Google in terms of backend operations.

The launch of Gemini 1.5 Flash highlights its status as a “distilled” and highly efficient version of Gemini 1.5 Pro, optimized for “narrow” and “high-frequency” generative AI tasks. Given the substantial costs associated with maintaining a chatbot platform, similar to OpenAI’s ChatGPT, Google is likely eager to leverage cost-saving opportunities that also improve performance.

Additionally, Google is expanding Gemini’s context window to 32,000 tokens, which is about 24,000 words or 48 pages of text. The context window refers to the amount of input data a model uses to generate output. A larger context window enhances the model’s ability to summarize and analyze longer texts more effectively and reduces the likelihood of forgetting recent information in chatbot interactions.

Previously, file uploads for analysis required Gemini Advanced, the paid version available through Google’s $20-per-month Google One AI Premium Plan. However, Google has announced that soon all Gemini users will be able to upload files from Google Drive and local devices.

Amar Subramanya, VP of Engineering at Google, explained in a blog post with TechCrunch, “You’ll be able to upload your economics study guide and ask Gemini to create practice questions. Gemini will also soon be able to analyze data files, uncover insights, and visualize them through charts and graphics.”

To address concerns about hallucinations—where a generative AI model fabricates information—Google is previewing a feature that will display links to related web content below certain Gemini-generated answers. English-speaking Gemini users in select regions will see a “chip” icon at the end of a paragraph, providing links to websites or emails (with Gmail access permission) for further exploration.

This update follows previous issues where Google’s generative AI models produced inaccurate information, such as recommending non-toxic glue in pizza recipes or creating fake book reviews. Earlier this year, Google introduced a “double-check” feature in Gemini to highlight statements corroborated or contradicted by other online sources. The new related content links aim to enhance transparency about the sources Gemini might be referencing.

The effectiveness and accuracy of these related links are yet to be determined.

In addition to these updates, Google is expanding Gemini’s reach. After an initial launch on select devices earlier this year, Gemini is now becoming available in the European Economic Area (EEA), U.K., and Switzerland, with new language options including French, Polish, and Spanish. Users can access Gemini in Messages by tapping the “Start chat” button and selecting Gemini as a chat partner.

The Gemini mobile app is also launching in more countries, and Google is extending access to teenagers globally. Previously available only in select countries, Gemini will now be accessible to teens in all regions where it is offered to adults. Google is implementing “additional policies and safeguards” for teen users, though specifics are not provided. A new onboarding process and an “AI literacy guide” will help teens use AI responsibly.

The ongoing debate about whether children are using generative AI tools appropriately or misusing them continues. Google aims to avoid negative publicity, such as accusations of Gemini being a tool for plagiarism or offering poor personal advice, and is taking steps to mitigate these risks.