Driven by its "AI-first" strategy, Meta is breaking down technological barriers—this social media giant, once proud of its Llama open-source model, now not only allows employees to widely use AI tools from competitors such as Google, OpenAI, and Anthropic, but also deeply integrates them into the entire daily workflow, from coding, writing, to design and decision-making.
According to internal documents disclosed by Business Insider, Meta employees can now legally access top external models including OpenAI's ChatGPT-5, Google's Gemini 3 Pro, and Claude (via an internal tool called Devmate). In June, Chief Information Officer Atish Banerjee clearly stated in a company-wide memo: "Making AI the core of our work style" is one of the company's top priorities, and emphasized "not being limited to self-developed technology, but focusing on practical results."
Meta's internal AI tool matrix has formed a "hybrid intelligence" ecosystem:
- Agentic Auto: A top-level agent that supports tool calling and collaboration with sub-agents. Integrated with GPT-5 Thinking, it can extend reasoning time for better answers;
- iLlama: An internal fast model optimized based on Llama 3.1, specifically designed to solve Meta's business problems;
- Llama 4 Maverick: An open version of the model without company-sensitive data;
- Devmate: A programming assistant that uses Claude at the underlying level;
- Metamate: A general internal AI assistant based on Llama.
More notably, Meta has officially incorporated Midjourney image generation technology into employee workflows. This tool was made available to all employees in October for "concept design and production," accelerating the development of creative prototypes. This marks that Meta has also chosen "open-source solutions" in the visual generation field, rather than fully relying on its own products like Vibes.
In terms of infrastructure, Meta has migrated its entire office suite to Google Workspace—email, documents, cloud storage, and chat have all been switched, with the reason being "unlocking AI-driven features" and seamless integration with tools such as Gemini and NotebookLM Pro. This move is essentially a clear stance in favor of Google in the cloud office war between Microsoft and Google.
To encourage AI application, Meta has also launched an internal gamification system called "Level Up": Employees who complete tasks using AI can earn badges, and performance evaluations will gradually include "AI-created value" metrics, with plans to implement them fully by 2026.
Ironically and practically, this company that has invested billions of dollars to build the Llama ecosystem now openly admits that a single technology stack cannot meet complex work scenarios. As Mahesh Sabharwal, a senior executive at "Reality Lab," said: "Our core is not about which model to use, but about giving employees the best intelligent experience."
