You found global talent. Deel’s here to help you onboard them

Deel’s simplified a whole planet’s worth of information. It’s time you got your hands on our international compliance handbook where you’ll learn about:

• Attracting global talent

• Labor laws to consider when hiring

• Processing international payroll on time

• Staying compliant with employment & tax laws abroad

With 150+ countries right at your fingertips, growing your team with Deel is easier than ever.

Get the guide

Today:

• Cohere Launches Aya Vision Model

• OpenAI Launches $50M Research Initiative

• Amazon Develops AI Reasoning Model

• Microsoft Launches Improved Copilot App

• Meta Unveils Brain2Qwerty Typing System

Cohere Launches Aya Vision Model

Cohere For AI has launched Aya Vision, a state-of-the-art multilingual, multimodal AI model. It excels in tasks like image captioning, text generation, and visual question answering across 23 languages. Aya Vision outperforms larger models, emphasizing efficiency and accessibility. Open-weight models (8B and 32B) are available on Kaggle and Hugging Face. Cohere is also launching Aya Vision Benchmark for better multilingual evaluation, fostering global AI research collaboration.

OpenAI Launches $50M Research Initiative

OpenAI has launched NextGenAI, a $50M initiative uniting 15 top research institutions to advance AI in education and research. The program provides grants, compute funding, and API access to accelerate breakthroughs in healthcare, digital learning, and AI literacy. Universities like MIT, Harvard, and Oxford are leveraging AI for research, education, and digitization. NextGenAI strengthens academia-industry ties, ensuring AI’s benefits reach students, scholars, and the broader community.

Amazon Develops AI Reasoning Model

Amazon is developing an AI reasoning model under its Nova brand, expected to launch in June. Inspired by models like OpenAI’s o3-mini and DeepSeek’s R1, it will use a hybrid reasoning architecture for better accuracy in math and science. Amazon aims to make it cost-efficient compared to competitors, though pricing could be a challenge. The model was first introduced at the re:Invent conference last year.

Microsoft Launches Improved Copilot App

Microsoft is rolling out a new native Copilot app for Windows, replacing its previous web-based version. Built with XAML, it features an improved UI, a side panel, conversation history, and voice interaction. The design now aligns with Copilot for macOS, supporting image uploads and text generation. The update is first available to Windows Insiders via the Microsoft Store and will expand to all Windows 11 PCs soon.

Meta Unveils Brain2Qwerty Typing System

Meta has introduced Brain2Qwerty, a non-invasive AI system that decodes brain activity into text while users type. Using EEG and MEG, the model achieves a 19% character error rate for the best participants, significantly improving over previous non-invasive methods. The study suggests decoding relies on both motor processes and cognitive factors, closing the gap between invasive and non-invasive brain-computer interfaces. This breakthrough enhances communication for non-verbal individuals.

🧠RESEARCH

Visual-RFT: Visual Reinforcement Fine-Tuning

Visual-RFT extends Reinforcement Fine-Tuning (RFT) to vision tasks, using Large Vision-Language Models to generate multiple responses and refine them with verifiable rewards. It significantly improves accuracy in few-shot classification and object detection, outperforming Supervised Fine-Tuning. This method enhances model adaptability and efficiency with minimal labeled data.

Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs

Phi-4-Mini is a 3.8B parameter model excelling in math, coding, and multilingual tasks with optimized data and expanded vocabulary. Phi-4-Multimodal integrates text, vision, and speech using LoRA adapters for efficient inference. It outperforms larger models in vision-language and speech-language tasks, setting new benchmarks in compact AI performance.

Difix3D+: Improving 3D Reconstructions with Single-Step Diffusion Models

Difix3D+ enhances 3D reconstruction and novel-view synthesis using a single-step diffusion model, Difix, to remove artifacts and improve rendering quality. It refines underconstrained regions during training and inference, boosting realism across NeRF and 3D Gaussian Splatting. The approach doubles FID score improvements while preserving 3D consistency.

🛠️TOP TOOLS

Logo Redesign AI - AI-powered tool that transform and enhance your brand’s visual identity with ease and efficiency.

Copyleaks AI - Designed to identify and flag text generated by AI models.

Figurative Language Checker - Scans your text and ultimately identifies figurative language elements like similes, metaphors, and personification.

Neural Frames - Audioreactive AI animations for musicians, creatives and visual artists.

Augie AI - Designed to make video-first marketing accessible and intuitive for businesses of all sizes.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Amazon's AWS has launched a new group focused on agentic AI, led by Swami Sivasubramanian, to automate tasks and drive innovation. Internal restructuring aims to accelerate AI development, with shares slightly down at $204.81.

• Contextual AI's new grounded language model (GLM) surpasses GPT-4o in accuracy, scoring 88% on the FACTS benchmark. Designed for enterprise RAG applications, it minimizes hallucinations, optimizes retrieval, and integrates structured data for precise, trustworthy AI responses.

• The LA Times is using AI to label opinionated articles and generate "Insights," but errors have sparked backlash. Critics argue the AI lacks editorial oversight, producing misleading bias assessments and controversial historical interpretations.

• The U.S. government anticipates artificial general intelligence (AGI) within two to three years, raising national security, economic, and labor concerns. Former AI adviser Ben Buchanan warns of profound global impacts, urging preparation amid uncertainty and geopolitical competition.