Reimagining K-12 Education: How Blended Labs is raising the bar with AI-driven schools

Blended Labs, an educational technology and AI company based in Germany, is using Llama 3.1 as the foundation of its flagship product, Blended OS. An AI-native school operating system, Blended OS streamlines administrative tasks and delivers tailored educational experiences under a single platform that makes it easy to efficiently set up and manage traditional and online schools of all sizes.

What makes Blended stand out is SphereAI, a unique, multidimensional schooling AI framework powered by Llama 3.1 8B and 70B. SphereAI enables a wide range of AI-native flows for personalized learning pathways, real-time feedback, instant educational content generation, and social gamification, catering to a new generation of AI-driven schools.

“Our integration of Llama has transformed the educational experience for over 10,000 users, including students, teachers, administrators, and parents,” says Mostafa Naeem, CEO at Blended. “Existing edtech solutions predominantly mirror traditional practices in digital form. At Blended, we’re radically improving school operations, enhancing learning results, and saving time across the board—raising the bar for what’s achievable in education.”

Blended OS brings state-of-the-art features to nearly every aspect of the digital school environment for administrators, teachers, and students. For example, the Blended Teacher’s AI Productivity Suite saves teachers an average of 10 hours per week. It includes more than 40 AI tools that can generate worksheets, help plan lessons, give assessment feedback, automate grading, and analyze student performance.

“With Blended’s Software as a Service model, teachers transition from task executors to operators, delegating repetitive responsibilities like grade analysis, feedback, and personalized recommendations to agentic AI systems,” Naeem says. “This transformation extends to every stakeholder in the school ecosystem, empowering students, administrators, and educators to focus on what truly matters: learning and growth.”

The company is launching a new set of capabilities enabled by the Llama 3.2 11B vision model that introduces visual tutoring support. One standout feature is Offline Assignment Grading. Traditionally, teachers can spend countless hours grading paper-based assignments, logging results into spreadsheets, and reconciling them with online activity data. The new SphereAI capability changes this. Students simply upload images of homework, projects, or in-class activities, and Blended will automatically extract, structure, and co-grade the submissions. Naeem says this is set to save teachers an additional four hours per week.

For students, Llama is the intelligence backend behind Lumi, Blended's 24/7 AI co-teacher. Lumi focuses on giving personalized help for language learning and STEM subjects, adjusting to each student’s pace and needs, and improving their learning results. With real-time feedback and personalized recommendations, students have seen a 22% improvement in learning outcomes.

For school administrators, Llama provides strong AI-powered analytics and automation, making tasks like school admissions, academic performance reports, and tracking teacher engagement insights easier. These features improve decision-making and operational efficiency across the board.

Schools using Blended OS have reduced operational costs by 18% through paperless workflows, streamlined communication, and centralized stakeholder engagement. But Naeem says the impact goes beyond efficiency. In regions impacted by crises such as Sudan, Blended has supported over 1,000 students with access to quality online education through fully virtual schools, ensuring that learning never stops—even in the most challenging circumstances.

Blended OS handles hundreds of thousands of data transactions every day. Currently, it has over 25 different data sources linked to learning and school operations, like admissions, exams, reporting, communication, discussion rooms, and behavioral analysis. To improve how schools make decisions and understand context, the team designed a graph-powered extension to Llama. This enables each school to have its own custom, secure, and privacy-preserving generative and analytical tools, contextualized using their own knowledge bases.

In addition, the team has developed multiple Llama-powered agentic workflows for education-focused tasks. For example, Blended AI can create, recommend, and adapt material that matches teaching objectives and standards for British and American diplomas, such as the Common Core standards. The team built it using a seed data bank that teachers helped populate and tag, and then generated thousands of pieces of content that fit those educational standards, giving teachers ready-to-use material for daily assessments. Blended AI also provides gamified question generation capabilities, which gives teachers better recommendations to help improve learning outcomes for students struggling with specific topics.

“Making sure Llama was hosted and maintained efficiently was a priority, especially for schools with privacy needs and in the competitive education market,” says Naeem. Blended supports self-hosting for some clients and has also partnered with vendors like Groq offering Inference-as-a-Service, a budget-friendly, speedy, and scalable option.

Keeping Llama safe for classroom use was also critical. The team added Azure Content Safety for prompt protection and content filtering to make sure both teachers and students are safe in educational environments. Blended is also integrating Llama Guard Vision to safeguard vision-based features.

“All these efforts are setting us up nicely to provide safe, responsible AI solutions for K-12 education,” says Naeem.

“In the next two to three years, we forecast that AI-driven software will radically change education— observing interests, identifying learning styles, and tailoring content with unparalleled precision,” says Naeem. “This is not a theoretical possibility; it’s a rapidly approaching reality.”

Open source AI will play an important role in that transformation, Naeem says, giving organizations the power to take back control of their technology.

“For companies like ours, open source means owning and shaping our intellectual property—whether it’s data, cognitive workflows, or highly specialized proprietary models,” he says. “This level of control allows us to align technology directly with our mission and set the pace for impactful innovation.”