DeepSeek gains token share as Anthropic dominates AI spending
Low-cost AI models gain traction in production workflows while Anthropic maintains spending leadership.
On a report by Vercel
Beat
Models
Small AI models enable diverse multi-agent financial simulation
Hugging Face Blog details how distinct small models power interactive finance game dynamics.
On a report by Hugging Face Blog
Small AI model simulates woodland economy with designed scarcity
Hugging Face Blog details a multi-agent economy experiment using a 3B model to simulate trade and market dynamics.
On a report by Hugging Face Blog
AIpressr
Industry Leaders Trust AIpressr to Move Their News
The press wire for the AI ecosystem — read by the journalists, analysts, and policy desks covering AI, and indexed by the models they use.
Submit a release →
Nemotron 3.5 enhances multimodal safety for enterprise AI systems
Hugging Face's latest update introduces customizable safety policies and multilingual support for AI content moderation.
On a report by Hugging Face Blog
Nemotron 3.5 ASR aims to streamline multilingual speech recognition
Hugging Face Blog outlines how Nemotron 3.5 ASR addresses common multilingual speech recognition challenges with a single model.
On a report by Hugging Face Blog
Hugging Face enhances Nemotron models with synthetic Q&A data
Task-seeded synthetic Q&A generation improves Nemotron model performance across multiple benchmarks.
On a report by Hugging Face Blog
Nvidia’s Nemotron 3 Ultra launches on Vercel AI Gateway
Nvidia’s latest reasoning model targets multi-turn agent workflows with a 1M token context window.
On a report by Vercel
Sponsor — MeMakie
Chat with AI characters that remember you
MeMakie is an AI character playground — build the cast, drop them in a group chat, or cast them in a story. Characters carry memory, react in unscripted ways, and surprise you.
xAI releases Grok Imagine Video 1.5 for synchronized audio-video generation
xAI's latest model enhances video generation with improved audio, photorealism, and character consistency.
On a report by Vercel
Microsoft releases smaller AI models MAI-Thinking-1 and MAI-Code-1-Flash
Microsoft introduces two new text LLMs with lower parameter counts, targeting reasoning and coding tasks.
On a report by Simon Willison
Financial firms adopt unified AI models for transaction data
NVIDIA-backed transaction foundation models aim to streamline AI systems in financial services.
On a report by NVIDIA Blog
AI weather startup reportedly surpasses government forecasts
Windborne Systems claims its AI-driven WeatherMesh model outperforms traditional European weather predictions.
On a report by TechCrunch AI
JetBrains releases Mellum2 for efficient AI inference
The 12B-parameter Mixture-of-Experts model targets latency-sensitive text-and-code workloads.
On a report by Hugging Face Blog
Alibaba's Qwen 3.7 Plus integrates vision and language on Vercel
Alibaba's latest AI model combines vision and language capabilities, now accessible via Vercel's AI Gateway.
On a report by Vercel
Simon Willison highlights AI cost trends and model updates
Simon Willison's May newsletter discusses rising AI costs, Anthropic's progress, and recent model releases.
On a report by Simon Willison
NVIDIA launches Cosmos 3 for unified physical AI reasoning and generation
New model combines reasoning and generation in a single architecture for robotics, autonomous vehicles, and simulations.
On a report by Hugging Face Blog
MiniMax M3 integrates with Vercel AI Gateway for multimodal AI
MiniMax's latest model, M3, offers a 1M-token context window and multimodal capabilities via Vercel AI Gateway.
On a report by Vercel