Learn AI in 5 Minutes a Day

AI Tool Report is one of the fastest-growing and most respected newsletters in the world, with over 550,000 readers from companies like OpenAI, Nvidia, Meta, Microsoft, and more.

Our research team spends hundreds of hours a week summarizing the latest news, and finding you the best opportunities to save time and earn more using AI.

Sign up with 1-Click

Today:

• Claude 3.5 Joins GitHub Copilot!

• Gemini Models on Copilot 

• OpenAI Adds AMD to Hardware

• xAI Eyes $40B Valuation

• Runway AI Streamlines Sales Process

• Meta’s Llama 4 Targets AMI

Claude 3.5 Joins GitHub Copilot

Claude 3.5 Sonnet is now integrated into GitHub Copilot, offering developers enhanced coding support directly within Visual Studio Code and GitHub.com. This model excels at solving real-world issues, achieving top scores on benchmarks like SWE-bench Verified and HumanEval for Python functions. Key features include code generation, debugging assistance, automatic test creation, and code explanations. 

Available in public preview, Claude 3.5 Sonnet operates via Amazon Bedrock for reliable performance. Over the coming weeks, it will be accessible to all GitHub Copilot Chat users and organizations.

Gemini Models on Copilot 

Google Cloud and GitHub have partnered to integrate Google’s Gemini models, starting with Gemini 1.5 Pro, into GitHub Copilot. This model offers developers multimodal capabilities and an extended context window of up to two million tokens, enabling it to handle extensive codebases and provide detailed insights. Gemini 1.5 Pro excels in code generation, analysis, and optimization, enhancing flexibility for developers to choose the best model for specific needs. 

Available soon on GitHub Copilot Chat and across various developer platforms, Gemini brings Google’s advanced code-assist features directly into popular IDEs, including Visual Studio Code and JetBrains.

OpenAI Adds AMD to Hardware

OpenAI is shifting its hardware strategy by incorporating AMD chips through Microsoft Azure to support ChatGPT and other projects, alongside Nvidia GPUs. Additionally, OpenAI is collaborating with Broadcom to develop custom AI chips, with production potentially starting in 2026, as it builds a dedicated chip team. 

Although plans for an OpenAI-specific foundry network were delayed due to cost, this new strategy positions OpenAI to compete with major tech players like Google, Microsoft, and Amazon, who have already invested heavily in custom AI hardware to manage operational costs and expand AI capabilities.

xAI Eyes $40B Valuation

Elon Musk’s AI company, xAI, is seeking to raise billions in funding, potentially boosting its valuation to $40 billion, nearly double its current $24 billion valuation. The new round would add to the $6 billion Series B raised earlier this year. 

xAI, which recently launched its AI model Grok, is training future models on a large Nvidia GPU cluster, aiming to leverage data from X (formerly Twitter) for competitive advantage. xAI has also introduced an enterprise API and utilizes X’s data to further train its models, following a recent policy change on data usage.

Runway AI Streamlines Sales Process

Finance software startup Runway developed an AI tool to enhance their sales process. The software scans potential customers' websites and drafts personalized emails based on the information it finds. 

To make these AI-generated emails appear more human, CEO Siqi Chen instructed the AI to include at least one typo in each message. Built over a single weekend, this tool saves the 12-person sales team dozens of hours weekly, allowing them to focus on engaging with interested clients. 

Meta’s Llama 4 Targets AMI

Meta is advancing AI with Llama, aiming to develop models capable of autonomous machine intelligence (AMI). The company’s upcoming Llama 4, expected in 2025, will enhance decision-making and adaptability in real-world scenarios. 

Meta leverages self-supervised learning (SSL) to build a broad data understanding, while reinforcement learning with human feedback (RLHF) refines specific tasks. This dual approach allows Llama to handle diverse tasks, from synthetic data generation for underserved languages to real-time problem-solving. 

🧠RESEARCH

Relaxed Recursive Transformers: Effective Parameter Sharing with Layer-wise LoRA

Recursive Transformers, a method to reduce the size and cost of large language models by reusing the same layer multiple times with minimal performance loss. Enhanced by Relaxed Recursive Transformers, which allow flexible adjustments, this approach outperforms similar models and improves processing speed by up to three times.

O1 Replication Journey: A Strategic Progress Report -- Part 1

The O1 Replication Journey introduces a novel approach to AI research replication through real-time documentation, fostering open science and rapid collaboration. Their “journey learning” model emphasizes process transparency, learning through trial and error, and outperformed standard methods on the MATH dataset by over 8%, showcasing AI’s potential in exploratory learning.

GPT-4o System Card

The GPT-4o System Card details an advanced multimodal model that handles text, audio, image, and video inputs, generating responses across these formats. GPT-4o is optimized for speed, delivering responses at near-human conversational rates, with 50% reduced API costs. Enhanced in non-English, visual, and audio tasks, it emphasizes safety through comprehensive evaluations and third-party assessments on societal impacts and potential risks.

Bielik 7B v0.1: A Polish Language Model -- Development, Insights, and Evaluation

Bielik 7B v0.1 is a 7-billion-parameter model designed for Polish language tasks, utilizing innovative training methods like Weighted Instruction Cross-Entropy Loss and Adaptive Learning Rate. It sets a new benchmark for Polish AI with substantial improvements over similar models, excelling in reasoning and role-playing on Polish NLP evaluation frameworks.

AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant

AgentStore is a platform for integrating diverse digital agents to automate complex computer tasks, combining generalist and specialist capabilities. With a MetaAgent core and AgentToken strategy, it adapts to various operating systems. Tests reveal significant improvements, more than doubling previous performance on benchmarks, showcasing its potential as a versatile computer assistant system.

🛠️TOP TOOLS

AImreply - AI email writing tool for transforming the content creation process for email.

APIPark - Open-source, all-in-one AI gateway and API developer portal, helping developers and enterprises easily manage, integrate, and deploy AI services

AIxBlock - A comprehensive platform to Productize AI models using decentralized computing resources

SwarmStack - AI agent for product planning 

Figflow - Translate designs into user stories.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Apple is launching new M4 Macs, reporting earnings, and expanding AI features, including Genmoji and ChatGPT integration. These updates mark Apple's strongest move yet in AI, aiming to enhance devices with smarter, creative capabilities.

• ElevenLabs acquired Omnivore’s team, creators of an open-source read-it-later app, to enhance its AI-powered reader, ElevenReader. With this acquisition, Omnivore’s tools will likely be integrated, enhancing ElevenReader's accessibility for readers.

• Oracle unveiled an AI-powered electronic health record (EHR) aimed at streamlining clinician workflows and reducing time spent on documentation. This new EHR integrates cloud technology and voice commands, improving data access and enhancing patient care.

• Sapien, a startup creating AI "coworkers" for finance professionals, secured $8.7 million in funding led by General Catalyst. Their tool helps CFOs streamline complex financial analysis tasks, enhancing productivity and catching critical errors faster.

• Crusoe Energy raised $500 million in funding, led by Peter Thiel’s Founders Fund, to support its AI-focused cloud infrastructure. The investment aims to scale Crusoe's environmentally friendly technology that repurposes wasted energy for cloud computing.