OpenAI releases o3-mini

Hey there 👋

We hope you're excited to discover what's new and trending in AI, ML, and data science.

Here is your 5-minute pulse...

print("News & Trends")

OpenAI releases its new o3-mini reasoning model for free (3 min. read)

OpenAI introduces O3 Mini, its most powerful small model yet—fast, cost-effective, and outperforming GPT-3.5 on key benchmarks. While powerful, it raises safety concerns with increased jailbreak risks and autonomy potential. OpenAI’s deliberative alignment mitigates risks, but challenges persist in balancing innovation, affordability, and control.

AI systems with ‘unacceptable risk’ are now banned in the EU (5 min. read)

The EU’s AI Act is now enforcing its first compliance deadline, banning "unacceptable risk" AI systems like social scoring and real-time biometric surveillance. With hefty fines looming, companies must navigate overlapping regulations while awaiting clearer guidelines. The next major enforcement phase kicks in August.

OpenAI unveils a new ChatGPT agent for ‘deep research’ (5 min. read)

OpenAI’s new "Deep Research" tool for ChatGPT Pro users promises precise, citation-rich answers for complex queries in fields like finance, science, and engineering. Powered by its o3 reasoning model, it tackles in-depth research but faces accuracy challenges. Currently web-only, it’s expanding soon with enhanced features and broader access.

Mistral AI releases Mistral Small 3, an open source mid-sized model, achieving low latency on local hardware (4 min. read)

Mistral AI’s new 24B-parameter model delivers low-latency inference, processing 150 tokens/sec with a 32K context window. It outperforms models 3x larger, runs on local hardware (RTX 4090, MacBook 32GB RAM), and is ideal for real-time AI applications.

print("Applications & Insights")

How to deploy and fine-tune DeepSeek models on AWS
DeepSeek-R1, the groundbreaking open-source reasoning model, is now deployable on AWS via Hugging Face! From Inference Endpoints to SageMaker and EC2, developers can fine-tune and scale these models seamlessly for advanced AI tasks like coding and logic. Dive in and supercharge your applications!

How to Run Parallel Time Series Analysis with Dask
Boost your time series analysis with Dask! This step-by-step guide shows how to leverage parallel computing for scalable, efficient analysis using Dask DataFrame. Perfect for handling large datasets while keeping performance high.

Building effective agents with LangGraph
Master five essential workflow and agent patterns with LangGraph, inspired by Anthropic's "Building Effective Agents" blog. Learn to construct prompt chaining, parallelization, routing, orchestrator-worker, and evaluator-optimizer workflows from the ground up. Gain clarity on when to use workflows versus agents, refine execution flow, and harness LangGraph’s strengths for structured, scalable agent development.

How to Summarize Scientific Papers Using the BART Model with Hugging Face Transformers
Learn to summarize papers with BART using Hugging Face Transformers in Python! A great side project to sharpen your NLP skills and explore Hugging Face's powerful tools. Requires PyTorch installation.

print("Tools & Resources")

TRENDING MODELS

Text Generation
mistralai/Mistral-Small-24B-Instruct-2501
⇧ 18.4k Downloads
This model by Mistral AI is designed for instruction-based text generation tasks. With 24 billion parameters, it excels in understanding and generating detailed responses based on user instructions.

Text Generation
unsloth/DeepSeek-R1-GGUF
⇧ 301k Downloads
DeepSeek-R1-GGUF is a variant of the DeepSeek-R1 model, optimized for specific use cases requiring enhanced performance and efficiency. It maintains the core capabilities of the original model while offering improvements in speed and resource utilization.

Text Generation
m-a-p/YuE-s1-7B-anneal-en-cot
⇧ 19.1k Downloads
YuE-s1-7B-anneal-en-cot is a 7-billion-parameter model fine-tuned for English chain-of-thought reasoning tasks. It enhances logical reasoning and step-by-step problem-solving capabilities in text generation.

Image-Text-to-Text
Qwen/Qwen2.5-VL-7B-Instruct
⇧ 175k Downloads
Qwen2.5-VL-7B-Instruct is a vision-language model designed to process and generate text based on image inputs. With 7 billion parameters, it excels in tasks that require understanding and describing visual content.

Note: Multiple DeepSeek models such as the DeepSeek-R1 and Janus-Pro-7B are still trending. Not shown here to highlight other useful models.

TRENDING AI TOOLS

• 🤖 Kater: AI-powered insights from questions.

• 📌 SumoPPM: AI Solutions to increase business growth

• 🔄 Forloop: No-code platform for external data automation.

• 🔗 Merge: One API to handle hundreds of product integrations

That’s it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve the pulse experience for you.

See you soon,

Andres