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

Lab city, Japan

Tokyo

Sakana AI, Preferred Networks, NTT, SoftBank and national labs building Japanese-language models.

Tokyo is the centre of Japan's language model work. Sakana AI, founded in 2023 by former Google researchers David Ha and Llion Jones, a co-author of the transformer paper, works on new training methods; Preferred Networks builds the PLaMo models; NTT, SoftBank and NEC build their own; and the National Institute of Informatics leads LLM-jp, an open model project.

Builders

Japanese models include NTT's tsuzumi, launched in March 2024 as a light model that runs on a single GPU, SoftBank's Sarashina, the Swallow models from the Institute of Science Tokyo (formerly Tokyo Institute of Technology) and AIST, and Fugaku-LLM, trained in 2024 on the Fugaku supercomputer in Kobe. Sakana AI became known for evolutionary model merging, which combines existing open models into new ones without retraining.

Policy and foreign labs

The economy ministry's GENIAC programme, launched in 2024, gives model builders computing time. Japan's approach is light-touch: its 2025 AI Promotion Act relies on guidance rather than penalties, and its copyright law allows wide use of data for training. OpenAI opened its first Asian office in Tokyo in April 2024, and Anthropic opened one in 2025.

The weekly digest

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