Originally published at ssojet
Meta has officially released the first models in its new Llama 4 family—Scout and Maverick—marking a step forward in its open-weight large language model ecosystem. Designed with a native multimodal architecture and a mixture-of-experts (MoE) framework, these models aim to support a broader range of applications, from image understanding to long-context reasoning.
Llama 4 Scout includes 17 billion active parameters distributed across 16 experts, optimized to run on a single NVIDIA H100 GPU. It supports a 10 million token context window, making it suitable for general-purpose AI tasks. Llama 4 Maverick, also with 17 billion active parameters but utilizing 128 experts, provides enhanced capabilities in reasoning and coding, outperforming several models in its class based on Meta’s benchmarks.
Both models were distilled from Meta’s flagship model, Llama 4 Behemoth, which has 288 billion active parameters and nearly two trillion total. Meta claims the Behemoth surpasses GPT-4.5 and other competitors on multiple STEM benchmarks. Despite not being fully released, Behemoth serves as a key training teacher for the smaller Scout and Maverick models.
Beyond model architecture, Meta emphasized a revamped training and post-training strategy, including lightweight supervised fine-tuning, reinforcement learning, and a new curriculum design for handling multimodal input. These changes aim to improve performance across difficult tasks while maintaining efficiency and reducing model bias.
While benchmark numbers show the Llama 4 models performing competitively with industry leaders, some early users are expressing skepticism about their performance. Concerns about the models' capabilities have been shared on platforms like Reddit, indicating a potential disconnect between benchmark results and real-world application.
Llama 4 Scout and Maverick are now available for download on llama.com and Hugging Face.
Meta's Head of AI Research Announces Departure
Meta's Vice President of AI Research, Joelle Pineau, announced her departure from the company effective May 30. Her exit occurs as Meta seeks to strengthen its position in the AI market against competitors like OpenAI and Google. In her LinkedIn announcement, Pineau highlighted the importance of creating opportunities for others within the organization to pursue AI advancements.
Pineau has been a significant figure in Meta's AI research, overseeing the company's cutting-edge studies, including the development of the Llama family of AI models. Her departure comes just ahead of Meta's LlamaCon AI conference, where the company is expected to present new developments in its AI offerings, including the anticipated Llama 4.
Google Debuts Gemini 2.5 in the ‘Winner-Take-All’ AI Model Race
Google has introduced Gemini 2.5, its most advanced generative AI model, which reportedly outperforms models from OpenAI, Anthropic, Grok, and DeepSeek in various industry benchmarks. Gemini 2.5 is designed as a thinking or reasoning model, enhancing performance by analyzing information and drawing logical conclusions before responding.
This model is natively multimodal, allowing it to understand and process text, audio, video, images, and code, ensuring seamless integration into various applications. It boasts a context window of 1 million tokens, which is essential for handling complex queries and tasks, setting it apart from its competitors.
Gemini 2.5 is available to Gemini Advanced paid users and will soon be integrated into Google Cloud's Vertex AI platform, providing enterprises with a powerful tool for developing custom applications.
In the competitive landscape, models like Gemini 2.5 signify the ongoing race among AI developers to enhance capabilities and performance, which is crucial for businesses looking to leverage AI for their operations.
For companies focusing on security and user management, implementing secure Single Sign-On (SSO) and multi-factor authentication (MFA) through solutions like SSOJet's API-first platform can streamline user experiences while maintaining high security standards. Explore SSOJet's offerings at https://ssojet.com for secure authentication solutions tailored for enterprise clients.