Okay, here’s a revised and fact-checked version of the article, incorporating corrections and updated information based on web searches as of today, november 2, 2023. I’ve focused on verifying dates,financial figures,and company details.
—
## Cohere Launches New Language Models Optimized for On-Device Use
AI startup Cohere has announced a new family of language models designed to run directly on devices, even without an internet connection. These models were trained on a single cluster of 64 Nvidia H100 GPUs using relatively modest computing resources,making them accessible to a wider range of developers. Cohere emphasizes that the models are particularly well-suited for applications serving users who speak diverse native languages.
The company highlights that its underlying software is optimized for on-device usage, requiring less computational power then many comparable models. This capability is particularly valuable in countries with limited or unreliable internet access, such as India, where it can unlock a wide range of applications.
The models are now available on HuggingFace, the cohere Platform, Kaggle, and Ollama for local deployment. Cohere is also releasing the training and evaluation datasets on HuggingFace and intends to publish a technical report detailing its training methodology.
techcrunch event
boston, MA
|
June