EurekAlert!

MathEval: a comprehensive benchmark for evaluating large language models on mathematical reasoning capabilities

This study introduces MathEval, a comprehensive benchmarking framework designed to systematically evaluate the mathematical reasoning capabilities of large language models (LLMs). Addressing key ...

China's Z.ai GLM-5.2 tops OpenAI’s GPT 5.5 model on key benchmarks

Chinese startup Z.ai has launched GLM-5.2, a powerful AI model for complex coding projects. This new large language model boasts a massive 1 million token context window, allowing it to handle ...
Geeky Gadgets

AI Benchmarks Are Broken : The Leaderboard Illusion

What if the tools we trust to measure progress are actually holding us back? In the rapidly evolving world of large language models (LLMs), AI benchmarks and leaderboards have become the gold standard ...
ascopubs.org

RadOncRAG: A Novel Retrieval-Augmented Generation Framework Improves Large Language Model Benchmark Performance in Radiation Oncology

Large language models (LLMs) show promise in assisting knowledge-intensive fields such as oncology, where up-to-date information and multidisciplinary expertise are critical. Traditional LLMs risk ...

Z.ai’s open-weights GLM-5.2 beats GPT-5.5 on multiple long-horizon coding benchmarks for 1/6th the cost

It allows engineering teams to host frontier-level AI on their own sovereign infrastructure, entirely eliminating vendor lock ...

Small Language Models Outperform Frontier AI On Cost, Speed And Accuracy

Bigger has defined AI from day one. New data says task-specific small models beat frontier LLMs on accuracy, cost and speed — ...
The LancetOpinion

Deception in clinical large language models: an under-recognised safety risk

Large language models (LLMs) are rapidly being integrated into clinical workflows, supporting tasks such as diagnosis ...
PsyPost on MSN

Artificial intelligence models show massive gaps on traditional human intelligence tests

Artificial intelligence programs designed to process and generate text show remarkably high verbal reasoning abilities, but ...

Waterloo's PAW compiles task specs into 23MB LoRA adapters a 600M-parameter model runs entirely offline.

Local AI inference at 32B-parameter quality, no cloud API required: University of Waterloo researchers released PAW on July 2, 2026, a system that compiles any natural-language task spec into a 23MB ...