June 2025 has been a particularly good month for artificial intelligence. We have seen the release of new versions of language models, improvements in agent autonomy, key announcements from technology giants, and an increased presence of AI in everyday services and enterprise solutions.
In this article we discuss some of the main developments that are shaping the direction of this technology.
Advances in autonomy and specialization of AI models
GPT – 4.5 Turbo: faster, more economical and more useful
OpenAI has taken another step forward in its roadmap with the release of GPT-4.5 Turbo, an enhanced version of its flagship model. It is not as radical an architectural change as GPT-5, but it does represent a significant evolution in several key aspects:
- Higher speed and lower cost per token, allowing integration into more products.
- Extended context up to 1 million tokens for enterprise users, facilitating long-running projects without fragmentation.
- More robust reasoning, especially in data analysis, planning and coding tasks.
GPT – 4.5 Turbo is now available on ChatGPT and has been rapidly adopted by developers of productivity applications, virtual assistants and business automation solutions. Tests indicate that it can maintain session context for days and execute instructions with an autonomy that previously required constant human intervention.
Devin, reaches basic professional level
Cognition Labs has updated its Devin agent, designed to program, test, debug and maintain software autonomously. In this new version, Devin has moved beyond tasks that previously required a junior developer:
- Solves bugs without the need for step-by-step explanation.
- Integrate new libraries and frameworks from technical documentation.
- Manage complex projects with multiple files, versions and dependencies.
What is most remarkable is that Devin can operate for days on end, learning from his own mistakes and adjusting his behavior as a professional trainee would. Some startups are already incorporating him as a “virtual” member of their development teams.
Integration of AI into everyday products and services
Apple makes a strong entrance with “Apple Intelligence” at WWDC 2025
The company introduced Apple Intelligence, its comprehensive artificial intelligence system that will be progressively integrated into all devices:
- Siri completely revamped, now with generative capabilities, contextual understanding and multimodal response.
- Mail, Safari and Notes incorporate features such as automatic summaries, smart suggestions, and assisted writing.
- Simultaneous FaceTime translation, useful for both users and companies with international teams.
- Generation of images in apps such as Messages, Keynote and Freeform.
Apple also announced a collaboration with OpenAI, incorporating ChatGPT-4.5 Turbo as an optional engine within their systems. Changing its strategy of developing its own AI without relying on third parties.
Meta launches personalized AI in WhatsApp and Messenger
Meta, meanwhile, has taken another step towards personalization of its messaging platforms with the launch of Meta AI Agents, initially available in the U.S. and Canada.
Allowing users:
- Create customized virtual assistants, with visual identity, communication style and conversational memory.
- Automate frequent tasks such as purchasing, reservations, event planning or technical inquiries.
- Use agents with specific functions within group chats, such as moderators, translators or automatic summaries.
At the business level, this opens the door for SMEs and large companies to have virtual assistants integrated into WhatsApp Business, improving customer service, sales automation and loyalty.
Strategic alliances and industry: AI as a new corporate standard
NVIDIA and SAP: Partnership to Transform Enterprise AI
One of the biggest news of the month was the partnership between NVIDIA and SAP, two giants in their respective industries. The goal is to integrate advanced AI capabilities into SAP’s enterprise resource planning (ERP) systems, both in its S/4HANA version and in SAP Business One, which is very popular among SMEs.
This collaboration will allow:
- Automate processes such as accounting, demand forecasting or logistics.
- Incorporate real-time virtual assistants for each department (finance, HR, purchasing…).
- Run predictive analytics with high-performance GPUs, accelerating response times.
This alliance reinforces the trend of bringing AI to key business processes, transforming traditional tools into intelligent ecosystems.
Huawei launches PanGu Chat 2.0: China’s answer to ChatGPT
Huawei has updated its PanGu Chat language model, which is already positioned as the national alternative to GPT-4 within the Chinese ecosystem. Version 2.0, introduced in mid-June, incorporates notable improvements in:
- Comprehension of technical and scientific texts.
- Responsiveness based on legal, medical and academic databases.
- Generation of code and solutions for regulated business environments.
Although its availability is restricted to the Chinese market due to regulatory issues, PanGu Chat represents an important step in China’s technological sovereignty, and in the specialization of models for specific sectors.
AI, regulation and the immediate future
Regulation: EU accelerates implementation of the AI Act
Although it was passed at the end of 2024, the European AI Act has entered an intensive enforcement phase since June. Tech companies are being notified about new obligations on transparency, risk control and oversight of high-impact algorithms.
OpenAI, Google, Meta and Anthropic have started to adapt their models to show traceability in their answers, as well as warnings about possible biases or limitations. This has also spurred the development of European open source models, such as those powered by Mistral AI and Aleph Alpha.
Training and business adoption
Finally, more and more companies are investing in artificial intelligence training for their employees, aware that the competitive advantage no longer lies only in having access to technology, but in knowing how to use it judiciously and effectively.
Tools such as Flowtask, Microsoft Copilot or Notion AI are being massively adopted for everyday tasks, while innovation departments explore customized models adapted to their internal data and flows.
What were once spectacular demonstrations in laboratories are now part of everyday experience. June 2025 has cemented AI’s definitive move from the realm of emerging innovation to basic infrastructure.
Faster models, more autonomous agents, deeper integrations and strategic decisions that will affect the coming years reflect a time of structural change in the way we work, communicate and make decisions.
Artificial intelligence is no longer optional. It is the new standard.
