I’ll do AI related announcements first.
Meta has announced the release of its newest AI models, Llama 4, which now enhance the Meta AI assistant across platforms like WhatsApp, Messenger, and Instagram. The two new models, Llama 4 Scout and Llama 4 Maverick, offer impressive capabilities; Scout can operate within a single Nvidia H100 GPU and has a context window of ten million tokens. Meta claims that it outperforms Google’s Gemini 3 and other competitors in various benchmarks. Meanwhile, the larger Maverick model shows comparable performance to OpenAI’s GPT-4o. Meta is also training a third model, Llama 4 Behemoth, which boasts 288 billion active parameters and aims to excel in STEM benchmarks. The company has transitioned to a “mixture of experts” architecture for Llama 4, enhancing efficiency by optimizing resource use. However, Meta’s licensing restrictions have drawn criticism, with some arguing that it limits the open-source nature of the models.
Google has launched Gemini 2.5 Pro, its most advanced artificial intelligence model to date, which is now available at a competitive price of one dollar and twenty-five cents per million input tokens and ten dollars per million output tokens. The model, designed for developers, aims to increase accessibility and has drawn significant interest since its release. Google has responded to this enthusiasm by raising rate limits and moving Gemini 2.5 Pro into public preview in the Gemini API. Developers using the model during this preview phase will benefit from increased limits, while the experimental version remains free but with reduced capabilities.
Microsoft has unveiled significant updates to its AI assistant Copilot. The new features include memory capabilities that allow Copilot to remember user preferences and interests, enhancing personalization. Users can now tailor their interaction with Copilot, even opting for nostalgic appearances like Clippy. Additionally, the assistant will gain the ability to perform tasks online, such as booking tickets and making purchases. The updates also introduce Copilot Vision, enabling the AI to analyze content from screens and camera feeds.
Why do we care?
For IT service providers, the challenge lies in choosing the right tools that align with client needs while balancing performance, cost, and compliance. Strategic adoption requires not just technical integration but also risk assessmentrelated to licensing, pricing, and data privacy. And the insight from yesterday from Microsoft – there’s more usecase value staying six months behind the frontier models.
