Meta Platforms on Saturday released the latest version of its large language model (LLM) Llama, called the Llama 4 Scout and Llama 4 Maverick, along with a preview of Llama 4 Behemoth. These two new models are now available for download from Meta and Hugging Face. Meta says it’s still in the process of training Llama 4 Behemoth. Meta CEO Mark Zuckerberg, in an Instagram post, said two more models are on the way: one is Llama 4 Reasoning and the other is Llama 4 Behemoth, which he calls “the highest performing base model in the world.”
Meta added that Llama 4 Maverick and Llama 4 Scout will be open-source software. “We’re introducing Llama 4 Scout and Llama 4 Maverick, the first open-weight natively multimodal models with unprecedented context length support and our first built using a mixture-of-experts (MoE) architecture. We’re also previewing Llama 4 Behemoth, one of the smartest LLMs in the world and our most powerful yet to serve as a teacher for our new models,” Meta said in a blog post on April 5.
According to Meta, Llama 4 Scout has a 10-million-token context window — the working memory of an AI model — and beats Google’s Gemma 3 and Gemini 2.0 Flash-Lite models, as well as the open-source Mistral 3.1, “across a broad range of widely reported benchmarks,” while still “fitting in a single Nvidia H100 GPU.” Meta makes similar claims about its larger Maverick model’s performance versus OpenAI’s GPT-4o and Google’s Gemini 2.0 Flash, and says its results are comparable to DeepSeek-V3 on coding and reasoning tasks using “less than half the active parameters.”
Meanwhile, Llama 4 Behemoth has 288 billion active parameters with 2 trillion parameters in total. While it hasn’t been released yet, Meta says Behemoth can outperform its competitors (in this case GPT-4.5 and Claude Sonnet 3.7) “on several STEM benchmarks.”
For Llama 4, Meta says it switched to a “mixture of experts” (MoE) architecture, an approach that conserves resources by using only the parts of a model that are needed for a given task. As a general purpose LLM, Llama 4 Maverick contains 17 billion active parameters, 128 experts, and 400 billion total parameters, while the smaller model, Llama 4 Scout, is a general purpose model with 17 billion active parameters, 16 experts, and 109 billion total parameters.
Llama 4 Scout dramatically increases the supported context length from 128K in Llama 3 to 10 million tokens. “This opens up a world of possibilities, including multi-document summarisation, parsing extensive user activity for personalized tasks, and reasoning over vast codebases,” Meta said.
The company says that it trained both models on a wide variety of image and video frame stills in order to give them broad visual understanding, including temporal activities and related images. “These new models are important building blocks that will help enable the future of human connection,” Meta said.
Meta says Llama 4 models are already integrated into Meta AI experiences across WhatsApp, Messenger, Instagram Direct, and the Meta.ai website. Developers and enterprises can begin building with Scout and Maverick, with more details will be unveiled at LlamaCon on April 29.
Meta says this work was supported by its partners across the AI community, including Accenture, Amazon Web Services, AMD, Arm, CentML, Cerebras, Cloudflare, Databricks, Deepinfra, DeepLearning.AI, Dell, Deloitte, Fireworks AI, Google Cloud, Groq, Hugging Face, IBM Watsonx, Infosys, Intel, Kaggle, Mediatek, Microsoft Azure, Nebius, NVIDIA, ollama, Oracle Cloud, PwC, Qualcomm, Red Hat, SambaNova, Sarvam AI, Scale AI, Scaleway, Snowflake, TensorWave, Together AI, vLLM, Wipro.
Also Read: HCLTech Public Sector Arm; Infosys Formula E Stats Center; Accenture Drives AI Transformation
2. SandboxAQ Raises USD 450 Million in Series E
Enterprise quantitative AI company SandboxAQ has announced the close of its Series E funding round, securing over USD 450 million from major new investors including Ray Dalio, Horizon Kinetics, BNP Paribas, Google, and Nvidia. The round brings the company’s total funding to more than USD 950 million since its 2022 spin-out from Alphabet.
The fresh capital will accelerate the development and deployment of SandboxAQ’s Large Quantitative Models (LQMs), which are being used to solve complex scientific and quantitative challenges across industries such as biopharma, chemistry, materials science, cybersecurity, and financial services.
“The investments underscore the confidence in SandboxAQ’s vision and the transformative power of its Large Quantitative Models (LQMs), which are redefining how enterprises leverage AI to tackle complex scientific and quantitative challenges,” SandboxAQ said on April 4.
These new investors join a group backing SandboxAQ, including Breyer Capital, Eric Schmidt, Alger, Paladin Capital, S32, TIME Ventures, and funds and accounts advised by T Rowe Price Associates. The company noted that this expanded support strengthens SandboxAQ’s leadership in AI and quantum technologies, driving innovation across multiple industries.
“AI and advanced computing are having a powerful impact on financial services, and BNP Paribas Global Markets is proud to be at the forefront of this trend,” said Olivier Osty, BNP Paribas Head of Corporate and Institutional Banking Global Markets. “I look forward to working with SandboxAQ to explore innovative solutions at the crossroads of AI and quantum techniques.”
“I bet on the SandboxAQ team and its approach to Large Quantitative Models (LQMs) because I’m impressed by them both,” said Ray Dalio, Founder, Bridgewater Associates.
SandboxAQ is a B2B company delivering solutions at the intersection of AI and quantum techniques.
Also Read: Oracle AI Agent Studio, Deloitte Zora AI, Accenture AI Refinery Platform, NTT DATA Agentic AI Services
3. Isomorphic Labs Raises USD 600 Million to Accelerate AI-Driven Drug Discovery
Isomorphic Labs, an AI-first drug design and development company, has raised USD 600 million in its first external funding round, led by Thrive Capital with participation from GV and follow-on investment from parent company Alphabet.
Founded by Demis Hassabis, also CEO of Google DeepMind, Isomorphic Labs aims to leverage the power of AI to reimagine and accelerate drug discovery to bring much-needed treatments to millions of patients globally. The funding will accelerate the development of the company’s next-generation AI drug design engine, expand its internal programs into clinical stages, and scale its team of interdisciplinary experts.
“We’re excited to bring together a top-tier investor group with deep AI and life sciences expertise as we aim to transform this industry through an interdisciplinary approach,” said Isomorphic Labs founder and CEO, Demis Hassabis. “This funding will further turbocharge the development of our next-generation AI drug design engine, help us advance our own programs into clinical development, and is a significant step forward towards our mission of one day solving all disease with the help of AI.”
Isomorphic Labs has developed a suite of advanced AI models that form a unified platform for drug discovery across multiple therapeutic areas. An example breakthrough model includes AlphaFold 3, launched in collaboration with Google DeepMind in 2024, which can accurately predict the structure and interactions of all of life’s molecules.
The company is currently advancing both internal programs—primarily focused on oncology and immunology—and partnered initiatives with pharmaceutical companies such as Eli Lilly and Novartis. The latter recently expanded its collaboration with Isomorphic, over a year after its inception.
“At Thrive, we aim to invest in category-defining companies with singular conviction, and that has led us to Isomorphic Labs,” said Joshua Kushner, Founder and CEO of Thrive Capital. “We believe Isomorphic has earned a rare position to define a new age of drug discovery and design, and we are deeply inspired by their mission and the extraordinary progress they have made to date.”
“The investment will accelerate Isomorphic Labs’ frontier AI research and development, to rapidly advance the company’s next-generation AI drug design engine. Isomorphic Labs will consequently be expanding its team of world-class talent as part of its next phase of growth,” the official release said.
“AI and machine learning have long held the promise of transforming drug discovery, yet few companies over the past decade have unlocked their full potential. After witnessing the extraordinary pace of innovation at Isomorphic Labs, we believe their pioneering approach will redefine AI-powered drug discovery,” said Krishna Yeshwant, Managing Partner at GV.
Also Read: Wipro Sovereign AI, Capgemini Agentic AI, TCS Air New Zealand Partnership, Tech Mahindra PV Solution
4. Papa Johns and Google Cloud Partner to Deliver AI-Powered Pizza Experiences
Papa Johns and Google Cloud have announced an expanded multi-year partnership aimed at transforming the pizza maker’s ordering and delivery experience using the power of artificial intelligence (AI). Through the partnership, Papa Johns has created an innovation team, named PJX, that will leverage Google Cloud’s AI, data analytics and machine learning (ML) capabilities to provide a more frictionless and consistent experience for customers both inside its restaurants and throughout its digital channels, the companies said in a joint statement on April 3, 2025.
Through its partnership with Google Cloud, Papa Johns expects to drive growth through increased order frequency, higher-value orders, reduced customer service costs, and improved customer satisfaction (via a chatbot powered by Google Cloud’s technology).
“Our partnership with Google Cloud will enable us to take personalization to the next level. We’re not just reacting to orders – we’re anticipating our customers’ needs and proactively providing tailored recommendations and offers. This isn’t just about convenience; it’s about creating a truly joyful and personalized pizza experience that builds lasting loyalty,” said Todd Penegor, President and CEO, Papa Johns.
PJX will focus on using Google Cloud’s AI to improve customer support, drive advanced personalization, and build proactive experiences “from click to crust,” the company said.
Key projects for the team include:
- Proactive Order Suggestions: Using Google BigQuery, Vertex AI, and Gemini, Papa Johns will proactively suggest orders through push notifications or email, based on learned customer preferences and anticipated needs for upcoming occasions like birthdays or sporting events.
- Hyper-Personalized Loyalty Programs: By applying generative AI to customer data, Papa Johns will dynamically customize its loyalty platform and app experience, tailoring promotions, rewards, and content in real time.
- Predictive Ordering and Marketing: AI will drive personalized marketing campaigns and provide predictive ordering shortcuts to enhance customer convenience and engagement.
- AI-Powered Customer Support: Papa Johns plans to build an AI-powered chatbot that can handle common customer inquiries, escalating complex issues to live agents. The pizza maker will also incorporate AI-powered voice ordering via the app to further enhance convenience.
- Smarter Restaurant Operations: Papa Johns will transition to a Google Cloud-based point-of-sale (POS) system that lays the foundation for AI-driven dispatching, route optimization, and intelligent automation of key restaurant processes.
“Papa Johns has been at the forefront of technology innovation – and its innovation team understands that AI has the power to transform the customer experience into something truly special,” said Matt Renner, President, Global Revenue, Google Cloud. “Google Cloud’s portfolio of data, analytics, and AI capabilities is helping Papa Johns deliver proactive, hyper-personalized service that goes above and beyond for its customers, building relationships that will keep customers coming back for more.”
Papa Johns is co-headquartered in Atlanta, Georgia, and Louisville, Kentucky and is the world’s third-largest pizza delivery company, operating more than 6,000 restaurants across approximately 50 countries and territories.
If this article saved you time or helped you decide better, consider supporting our work.
Support Telecomtalk
