Top 20 AI Updates This Summer: Don’t Miss These Crucial Insights!

This summer of 2025 was anything but a break for AI players. From major announcements to minor improvements by entities like OpenAI, Mistral, Figma, and Slack, the AI world is moving forward at full speed. Here is a summary of the key announcements you shouldn’t have missed this summer!

Increasingly Advanced AI Models

Undoubtedly, the major highlight of summer 2025 was the global launch of GPT-5. The latest from OpenAI, this new version was meant to outperform its predecessors in all aspects. However, it met with some disappointment from users, to the extent that the company had to backtrack and reintegrate its previous model.

Meanwhile, competitors took the opportunity to enhance their existing models or develop new features. Anthropic, for example, rolled out Claude Opus 4.1, boasting improved overall capabilities. The startup also equipped Claude with a memory feature, allowing it to draw on past conversations. Google’s Gemini also received a memory upgrade for more personalized interactions, as well as a temporary chat feature for fleeting discussions.

Additionally, OpenAI introduced two new open-source models named gpt-oss, marking a first since 2019 for the creators of ChatGPT.

AI in Coding and Development

This area has recently become a playground for AI solution providers: coding and software development assistants. Everyone is getting involved in various capacities:

• Microsoft unveiled GitHub Spark, an online development environment that allows the creation of applications through natural language instructions. The goal is to transform simple coding assistance into a full collaborative space where AI supports the entire project from conception to deployment.
• Google launched Gemini CLI, a free and open-source AI agent that operates directly in the terminal, providing access to the Gemini 2.5 Pro models for code generation, explanations, or task automation.
• Google also made Jules, its asynchronous coding agent powered by Gemini 2.5 Pro, available to the public. It can handle development tasks (testing, bug fixes, new features…) in a cloud virtual machine without requiring the developer’s constant oversight.
• Figma ended the beta of Make, its coding assistant launched in the spring, now offering it for free (with some restrictions) to its users. The tool enables the creation of prototypes and applications from textual descriptions.

Further Diversification of Generative AI

Moving beyond the realm of chatbots and text generators, there were also some new developments in video and audio this summer. Starting with Midjourney, whose version 7 image generator did not achieve the anticipated success, likely because the startup is focusing on a broader project of generating real-time worlds for which it has developed its own video generator, now available.

Targeted more at professionals, Adobe’s Firefly AI video model saw some enhancements this summer, including the integration of Google’s Veo 3 model and a sound effects generator. Google’s NotebookLM, already known for its ability to generate audio podcasts, can now do the same in video format.

On the voice front, Meta AI deployed an automatic dubbing system for Facebook and Instagram Reels, while the French company Mistral unveiled Voxtral, an open-source model focused on audio, featuring voice recognition, transcription, and more.

Everyday AI: The Expansion of AI-Assisted Learning

For some everyday uses, models or platforms have seen notable improvements this summer. Leveraging Voxtral, Mistral’s Le Chat now boasts an advanced vocal mode similar to ChatGPT, as well as enhanced search capabilities, widely deployed among most competitors. For productivity, the professional platform Slack has been enhanced with new AI features, including summaries, translations, and search capabilities.

However, it’s in the realm of assisted learning that new battles are shaping up. OpenAI recently added a Study mode to ChatGPT, which offers pedagogical tracking accompanied by quizzes and reflections to help you improve your skills in selected topics. Just a few days later, Google launched a similar mode in Gemini, named Guided Learning, along with revision tools for students.

The New AI Browser War

Unchallenged until now, Chrome is seeing unprecedented competition in the web browsing arena. Starting with OpenAI, which is preparing to launch its own browser based on Chromium. This upcoming tool will focus on AI agents capable of performing tasks autonomously, with an interface centered around chat.

At the same time, Perplexity has officially deployed its Comet browser, currently available only to subscribers at $200 per month. Here again, AI is not just an add-on but the core of the product. Accessible via a sidebar, Comet Assistant can analyze a page, summarize emails, manage a calendar, or make reservations.

Lastly, Microsoft continues to bolster Edge with new AI components, aiming to capitalize on its native integration of Copilot. These initiatives reflect a broader movement, with web navigation becoming a strategic battleground for AI giants, each hoping to assert their vision of an AI-driven browser.

AI-Related Statistics from the Summer

• 3%: the percentage of users who pay for AI services, according to Menlo Ventures.
• x10: AI-generated traffic has increased tenfold over a year, per a study by Ahrefs.
• 19%: AI slows down developers by 19% in problem-solving, according to the METR Institute.
• 95%: only 5% of AI projects in enterprises lead to a concrete operational deployment, reveals MIT.
• 96%: an impressive figure from Metricool, indicating the percentage of social media professionals who use AI at work.

Find all our articles on AI!