H-FARM PROGRAMS ON AI

AI Super Power
The Professional Master to bring AI into your department

AI Super Power is the Professional Master by H-FARM Business School for managers who want AI inside their workflows, not just in a browser tab.

In 85 hours you design no-code automations and AI agents, with a faculty that includes Diego Pizzocaro and Alessandro Scandaletti of H-FARM AI.

Starts on 27 October 2026 on the H-FARM Campus in Roncade. Part-time and taught in Italian, with in-person weekends and online evening sessions, for managers, department heads and entrepreneurs ready to lead AI adoption. Last Early Bird seats available.

• Hands-on workshops on real business briefs: data analysis, autonomous agents, multi-agent orchestration, AI Act compliance and security.

• You leave with an AI agent trained on your company’s data, an adoption roadmap for your team and an H-FARM Business School Open Badge.

🤖 Meta ships Muse Spark Open for Apple Silicon 🤖
Meta's 816B open-weight model rolls out with native Apple silicon integration.

Meta released Muse Spark Open, an 816B-parameter open-weight multimodal model built to run locally on Apple silicon. The model requires two M5 Ultra Mac Studios to operate, making it the largest open-weight model purpose-built for Apple hardware. Beyond the technical specs, the release signals a notable thaw in the Meta–Apple relationship, with both companies apparently finding common ground around on-device AI capabilities. This matters because it expands access to frontier-class models outside the cloud, giving developers and researchers a local inference option with serious muscle. The move also positions Meta as a leader in open-weight multimodal AI, raising competitive pressure across the industry.

🧠 Microsoft's 2.8T-parameter MAI-Base-2 tops benchmarks 🧠
Microsoft's largest model outpaces all open-weight rivals in pre-training speed and quality.

Microsoft detailed the training of MAI-Base-2, a massive 2.8 trillion-parameter mixture-of-experts model with 80B active parameters per forward pass. According to the company, MAI-Base-2 surpassed every open-weight model in pre-training evaluations, a clear signal of competitive pressure in the frontier model race. Microsoft shared in-depth information about its data strategy, compute efficiency techniques, and enterprise integration plans, suggesting the model will underpin future Azure AI and Copilot offerings. The capability jump matters for enterprise customers weighing open-weight alternatives: Microsoft is now competing directly with Meta's Llama family and other large-scale open models while maintaining a clear path to commercial deployment.

⚡ Nvidia drops Nemotron 3.5 Super VL for agents ⚡
Nvidia's new multimodal model delivers faster reasoning with fewer output tokens.

Nvidia released Nemotron 3.5 Super VL, a multimodal reasoning model that scored 26.5 on the Artificial Analysis Intelligence Index while using fewer output tokens than competing models. Alongside the flagship release, Nvidia shipped the smaller Bolt 74B-A9B variant and NVFP4 quantized weights, expanding access across different hardware tiers. The efficiency gains matter for enterprise deployment pricing: fewer tokens means lower inference costs at scale. Nemotron 3.5 Super VL is designed to power specialized AI agents capable of processing both text and visual inputs, positioning Nvidia not just as the chip supplier but as a direct competitor in the application-layer model market.

INTERESTING TO KNOW

🗣️ OpenAI rolls out GPT-Live-1.1 in six languages 🗣️

The rollout of GPT-Live-1.1 brings new voice options in French, German, Hindi, Mexican Spanish, Korean, and Japanese with improved accent accuracy. Available across ChatGPT, Codex, and dots, the update is rolling out to all users. The multilingual expansion signals OpenAI's push for broader global adoption, a key distribution move as voice-first AI interfaces become increasingly competitive territory.

🎙️ Fish Audio launches Drama 3 with word-level control 🎙️

Fish Audio released Drama 3, a text-to-speech model with a pricing-friendly twist: users can describe exactly how every word should sound and selectively re-record individual words without redoing the full clip. This granular integration of user direction into the synthesis pipeline could be a replacement for traditional voice-over workflows, giving creators fine-grained control that most competing TTS tools lack.