Pixtral 12B: Mistral’s first multimodal model

French AI startup Mistral has released its first multimodal model, Pixtral 12B. In other words, it has 12 billion parameters and can process both images and text. It is based on Mistral’s existing text model Nemo 12B and is said to be able to answer questions about any number of images of any size. Pixtral …

Read more

Information about OpenAI’s mysterious “Strawberry” project

OpenAI’s mysterious “Strawberry” project could see the light of day in the next two weeks. This is according to a report by The Information. Market observers had high expectations. However, the information that has now become public has cooled the enthusiasm considerably. According to the report, Strawberry is a pure text model and specializes in …

Read more

Translation specialist Smartcat raises $43 million

Smartcat, a provider of AI-powered translation tools, has raised $43 million in a Series C funding round. The Boston-based company provides companies with tools and services to translate written and spoken content into around 280 languages. As founder and CEO Ivan Smolnikov told TechCrunch, Smartcat uses a “matching engine” to select appropriate AI models for …

Read more

Workspaces for Claude help manage AI

Anthropic is introducing “Workspaces”, a new feature for managing AI systems in enterprises. The startup enables the creation and control of multiple isolated environments for Claude AI implementations, Michael Nuñez reports on VentureBeat. Companies can now set spending and usage limits, group API keys, and control access through user roles. The feature addresses key challenges …

Read more

SuperNova is a new model for enterprise use

Arcee AI has introduced SuperNova, a customizable language model with 70 billion parameters for enterprises. It can be used in a company’s own infrastructure and be customized, as James Thomason reports at VentureBeat. SuperNova is based on Meta’s Llama 3.1-70B Instruct architecture and uses a novel retraining process. It aims to provide an alternative to …

Read more

DeepSeek V2.5 celebrated as new open source champion

DeepSeek-V2.5 is the new champion among open source AI models. DeepSeek itself is an offshoot of the Chinese hedge fund High-Flyer Capital Management. The new model combines natural language processing and programming capabilities into one seemingly powerful system. According to Carl Franzen of VentureBeat, DeepSeek-V2.5 outperforms its predecessors in almost all benchmarks. The model offers …

Read more

ServiceNow announces AI agents for enterprises

ServiceNow is introducing a library of customizable AI agents for businesses. The company is planning updates to its Now Assist AI platform that will allow customers to integrate AI agents into their workflows. As Emilia David reports for VentureBeat, companies can use the new Now Assist Skill Kit to develop their own prompts and skills …

Read more

DigitalEx gives companies insight into the cost of generative AI

DigitalEx, a provider of cloud cost management software, has launched a new solution for controlling the costs of generative AI. The tool provides organizations with a centralized view of AI-related costs across multiple platforms, including AWS Bedrock, Azure OpenAI, and OpenAI. As CEO Sundeep Goel told VentureBeat, the solution enables detailed cost allocation, financial management, …

Read more

Transfusion enables combined text and image models

A new method called Transfusion enables the training of models that can process and generate both text and images. As researchers from Meta and other institutions report, Transfusion combines prediction of the next token for text with diffusion for images in a single transformational model. Experiments have shown that this approach scales better than quantizing …

Read more

OLMoE is a completely open source MoE model

A new open source model called OLMoE has been released by the Allen Institute for AI (AI2) in collaboration with Contextual AI. As Emilia David reports for VentureBeat, the model aims to be both powerful and inexpensive. OLMoE uses a mixed-expert architecture with 7 billion parameters, of which only 1 billion are active per input …

Read more

×