😺 Google just gave away its best AI

2026-03-27 · Source: The Neuron · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

Four distinct open-source AI models have been released under the Apache 2.0 license, collectively covering a wide range of compute needs from edge devices to cloud agents. Google's Gemma 4, available in four sizes, includes an edge model for Raspberry Pi (under 1.5GB memory) and a 31B model ranking #3 among open models. PrismML introduced 1-bit Bonsai, an 8B model compressed to 1.15GB, achieving 44 tokens/second on an iPhone. H Company launched Holo3, a 10B active parameter computer-use agent that set a new desktop automation benchmark. Arcee AI released Trinity-Large-Thinking, a 400B reasoning model (13B active) that scores #2 on agentic benchmarks at 96% less cost than Claude Opus 4.6. This collective release signifies a major shift, making competitive AI accessible and deployable across the entire compute spectrum.

Key takeaway

For CTOs and VPs of Engineering evaluating AI infrastructure, this proliferation of open-source, Apache 2.0 licensed models fundamentally alters your strategy. You can now own and customize frontier AI weights for diverse applications, from mobile to cloud, at a fraction of proprietary costs, enabling greater control and innovation without vendor lock-in.

Key insights

Open-source AI models now provide competitive, cost-effective solutions across all compute scales, from mobile to data centers.

Principles

• Open licensing (Apache 2.0) removes enterprise adoption friction.
• Specialized AI models complement generalists for diverse tasks.

Method

Deploying a suite of open-source AI models (Gemma 4, Bonsai, Holo3, Trinity) allows for optimized performance and cost efficiency across various device types and computational requirements, from edge to cloud.

In practice

• Run Gemma 4 E2B on Android for offline AI assistance.
• Utilize Bonsai for efficient AI on mobile devices.
• Deploy Holo3 for desktop automation tasks.

Topics

• Google Gemma 4
• Open-Source AI Models
• Apache 2.0 Licensing
• Edge AI
• AI Agents

Code references

• imbue-ai/mngr
• instructkr/claw-code

Best for: CTO, VP of Engineering/Data, Investor, AI Engineer, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Google Just Gave Away Its Most Powerful AI for Free, and Nobody Fully Grasps What That Means — Google's Gemma 4 release under Apache 2.0 enables powerful local AI inference, challenging cloud-based models.
• The Future of AI Is Open and Proprietary — AI's future is a hybrid ecosystem of open and proprietary models, orchestrated into specialized systems.
• There has been a situation in AI — GLM52 marks the arrival of a true open-source frontier AI model, challenging the perceived "moat" of proprietary systems like GPT-5.5 and Opus 48.
• not much happened today — Open models are rapidly advancing in coding and privacy, while enterprise AI platforms are consolidating agentic capabilities and specialized hardware.
• Top AI GitHub Repositories in 2026 — The open-source AI ecosystem is rapidly maturing, driven by local AI, agentic capabilities, and powerful open models.
• Gemma 4 and what makes an open model succeed — Open model success hinges on ease of use and adaptability, not just benchmark scores.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Neuron.