ARTIFICIAL INTELLIGENCE
GPT-OSS:20B/120B: https://openai.com/index/gpt-oss-model-card/. The newest open-weights models by Open AI, blazingly fast on Nvidia hardware
Magistral by Mistral: https://mistral.ai/news/magistral. The first reasoning model, in two configurations, by Mistral
Towards General-Purpose Model-Free Reinforcement Learning - a paper by META on their new Mr. Q Reinforcement Learning algorithm
Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback - an Arxiv paper by Chinese researchers on a potential breakthrough promising to be extremely efficient
Deepseek R1 - the Arxiv paper by the Chinese AI company Deepseek on chain-of-thought/reasoning that is shacking Wall Street’s Mag 7
NVIDIA Jetson Orin Nano Super - the edge AI solution recommended by ECONOVA-AI to implement substainable edge AI solutions for SMEs
Llama 3.3 70B - a very capable model by Meta, available on Ollama, LM Studio and Huggingface
The Simple Macroeconomics of AI - by Daron Acemoglu
NATIONAL BUREAU OF ECONOMIC RESEARCH - Working Paper 32487 - May 24 - DOI 10.3386/w32487. Acemoglu’s paper on the impact of AI on the economy. We disagree with the Nobel Price winner’s conclusions but we are all in favor of free of speech.
Llama 3.1 - Huggingface page with memory requirements to run the full model or its quantized versions according to the size of the context window selected
Mistral NeMo - a capable model by EU’s MIstral AI perfect for local deployment
Mixture-of-Agents Enhances Large Language Model Capabilities
by Junlin Wang, Jue Wang, Ben Arithawatkun, Ce Zhang, James Zou - Together AI - June 2024
QWEN2 - Huggingface page with operating requirements
ORPO: Monolithic Preference Optimization without Reference Model
Jiwoo Hong KAIST AI {jiwoo_hong, noah.lee, thorne}@kaist.ac.kr
Noah Lee KAIST AI {jiwoo_hong, noah.lee, thorne}@kaist.ac.kr
James Thorne KAIST AI
Tii Falcon docs
Microsoft Phi3 docs
X.ai Grok docs
MemGPT docs