Arcee-SuperNova: Flagship 70B LLM 🚀, Chai-1: The AI Model Set to Revolutionize Drug Discovery ☕, Apple announces its new A18 and A18 Pro iPhone chips 📱

Meet Arcee-SuperNova: Flagship 70B LLM for Enterprises

Arcee AI has launched SuperNova, a new AI model designed to compete with large language models like those from OpenAI, focusing on enterprise use. SuperNova comes in two versions: the powerful SuperNova 70B (70 billion parameters) and the efficient SuperNova-Lite 8B (8 billion parameters), both fine-tuned for precision and reliability.

SuperNova’s development combines cutting-edge techniques like distillation (from a 405B-parameter model), model merging, and Arcee AI's EvolKit pipeline, enhancing its performance in instruction-following, mathematical reasoning, and data privacy. SuperNova stands out for offering enterprises full ownership, customization, and the ability to deploy it on their own infrastructure, ensuring data security and reducing costs.

Available through AWS Marketplace, API access, and OpenWebUI, SuperNova also supports Reinforcement Learning from Human Feedback (RLHF) for continuous improvement. This makes it an appealing alternative to proprietary AI models, offering scalability, cost-efficiency, and customization for businesses.

Read the full version Here!

Chai-1: The AI Model Set to Revolutionize Drug Discovery

Chai Discovery, an AI startup backed by OpenAI and Thrive Capital, launched Chai-1, a revolutionary AI model designed to accelerate drug discovery by predicting molecular structures. Founded by former researchers from OpenAI and Meta, Chai Discovery has raised $30 million to develop advanced AI models for pharmaceutical research.

Chai-1 excels in predicting the structures of various biochemical molecules, including proteins, DNA, RNA, and small molecules, surpassing benchmarks set by Google's AlphaFold. It can handle a wide range of molecules and chemical modifications, making it highly useful for drug development. The model’s ability to predict protein-ligand interactions without the need for Multiple Sequence Alignments (MSAs) enhances its real-world applicability, especially when data is incomplete.

Chai-1 is available for free through a web interface, with its code and tools accessible for non-commercial use to encourage innovation. Its performance in predicting molecular structures positions Chai-1 as a breakthrough tool in drug discovery, with the potential to accelerate the development of new treatments in biotechnology.

Read the full version Here!

Apple announces its new A18 and A18 Pro iPhone chips

Apple has introduced the new iPhone 16 and iPhone 16 Plus, featuring the advanced A18 chip. This new chip brings significant improvements in performance and efficiency, supporting demanding tasks like AAA games, while providing a substantial boost to battery life. Built on second-generation 3-nanometer technology, the A18 chip enhances mobile gaming with hardware-accelerated ray tracing and up to 5x higher frame rates. It also powers advanced AI and machine learning capabilities, allowing the Neural Engine to run models twice as fast as previous versions.

Key features of the A18 chip include a 30% faster 6-core CPU, a 40% faster 5-core GPU, and improved power efficiency. These advancements ensure smoother performance, longer battery life, and immersive experiences for both everyday tasks and intensive gaming.

Additionally, the A18 chip enables new camera functionalities such as Camera Control and spatial photo and video capture, while also supporting Apple's AI and privacy-focused technologies.

Nvidia Free Course: An Even Easier Introduction to CUDA

Course Outline
In this notebook you will learn the basics of writing massively parallel CUDA kernels to run on NVIDIA GPUs. By the time you complete it you will be able to:

• Launch massively parallel CUDA Kernels on an NVIDIA GPU
• Organize parallel thread execution for massive dataset sizes
• Manage memory between the CPU and GPU
• Profile your CUDA code to observe performance gains

Learning Objectives

• Launch massively parallel CUDA Kernels on an NVIDIA GPU

• Organize parallel thread execution for massive dataset sizes

• Manage memory between the CPU and GPU

• Profile your CUDA code to observe performance gains

Topics Covered

The following topics are covered in this course:
CUDA C++
nvcc
nvprof

⭐⭐⭐

LIST OF TOOLS

✍️ Grammarly: AI-powered writing assistant.

📝 ProWritingAid: Grammar checking and style suggestions.

🤖 ChatGPT: Content generation and idea brainstorming.

📚 Jasper: AI writing assistant for technical documentation.

🏢 Writer: Enterprise AI writing tool.

🔄 Quillbot: Paraphrasing and rephrasing text.

📑 Scribe: AI-powered step-by-step guide generation.

✨ Wordtune: AI-driven writing style improvement.

🌍 DeepL Write: Writing assistance and translation.

🧠 ContentBot: Content ideas and drafting.

🗂️ Notion with AI Features: Drafting and summarizing documentation.

🔎 Ginger Software: AI-based grammar and spell-checking.

DeepSeek-V2.5 Wins Praise as New Open-Source AI Leader
DeepSeek V2.5 has emerged as a top open-source AI model, outperforming proprietary models like GPT-4 in function calling. Its use of synthetic data enhances privacy and performance, making it ideal for sensitive environments. Available on platforms like Hugging Face, it promotes accessible advanced AI tools, driving innovation in fields like coding and data analysis.
Why it Matters: DeepSeek V2.5 leads in performance while remaining fully open-source, pushing the boundaries of AI accessibility.
Practical Insight: Open-source AI models can rival proprietary systems, offering robust functionality without licensing fees.

Cerebras is Taking on Nvidia with World’s Most Powerful AI Chip
Cerebras is challenging Nvidia with its Wafer Scale Engine (WSE) chips, offering up to 900,000 AI-optimized cores. Its latest WSE-3 outperforms Nvidia’s H100 GPU, delivering 125 petaflops of AI performance. Cerebras aims to disrupt the AI hardware market with faster, more cost-effective solutions for AI training and deployment through its cloud-based AI Model Studio and AI inference services.
Why it Matters: Cerebras’ WSE chips could challenge Nvidia’s dominance in AI hardware with better performance and affordability.
Practical Insight: Cerebras is a serious competitor, offering alternatives for companies looking to reduce AI processing costs.


SiMa.ai Launches Modalix to Bring Multimodal Gen AI Everywhere
SiMa.ai launched the MLSoC Modalix, a highly efficient 6-nanometer chip designed for multimodal AI processing. Backed by $70 million in funding, Modalix supports advanced models like CNNs and Transformers while maintaining industry-leading energy efficiency. It aims to revolutionize AI-driven applications in sectors like healthcare, robotics, and aerospace by processing multiple inputs like text, audio, and video.
Why it Matters: SiMa.ai’s chip is set to advance edge AI, making it more efficient and scalable across industries.
Practical Insight: Businesses in sectors requiring AI-driven, multimodal processing can benefit from SiMa.ai’s innovative solution.


OpenAI to Release 'Strawberry' for ChatGPT
OpenAI plans to launch "Strawberry," a reasoning-focused AI model for ChatGPT in the coming weeks. Unlike traditional models that respond instantly, Strawberry will take time to “think” before answering, enhancing its reasoning abilities. The first release will focus on text-based tasks.
Why it Matters: Strawberry’s reasoning-based approach could elevate AI interactions, making responses more thoughtful and accurate.
Practical Insight: Strawberry represents a shift toward deeper, more analytical AI models that prioritize quality over speed in responses.

👥 Connect & Feedback!

👉 Join Us:

📧 Advertise In Weekly AI News:

📧 Contact directly at [email protected]

😍 Share with your friends!