News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM

University and research, Canada

Montreal

Yoshua Bengio's Mila: neural language models, attention for translation and AI safety research.

Montreal's Mila, the Quebec Artificial Intelligence Institute founded by Yoshua Bengio in 1993, is one of the largest academic deep-learning centres in the world. Bengio's 2003 paper on a neural probabilistic language model is an ancestor of every model on this site, and Mila researchers introduced the attention mechanism for translation in 2014.

Ideas from Mila

Bengio, who shared the 2018 Turing Award with Geoffrey Hinton and Yann LeCun, showed in 2003 that a neural network could learn word representations while predicting the next word. In 2014 Dzmitry Bahdanau, Kyunghyun Cho and Bengio added attention to neural translation, letting the model look back at the relevant source words; the transformer of 2017 built on this idea. Mila now gathers more than a thousand researchers from Montreal's universities.

Safety and industry

Bengio has become a leading voice on AI risk and chaired the International AI Safety Report, published in January 2025 by experts from 30 countries. In 2025 he launched LawZero, a non-profit working on AI that is safe by design. Google, Meta, Microsoft and Samsung keep research labs in the city, and ServiceNow, which bought Montreal's Element AI, co-led the BigCode project behind the StarCoder code models.

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.