. . AI: New Image Model Evaluation Arena and Snowflake Launches Arctic LLM (4.24.24)

Friends, in today's AI news, we see a new arena to evaluate models and get a first peek at the reigning champions. Additionally, a new model from the Snowflake AI Research Team has been launched across multiple platforms. Headlines from OpenAI and Anthropic on AI controls and safety have also been captured, alongside Apple's early steps in joining the open source movement.

Thanks for reading and sharing this newsletter! Now here’s today’s top AI news.

-Marshall Kirkpatrick, Editor

Snowflake AI Launches Cost-Effective LLM, Arctic

First impacted: AI researchers, Software developers
Time to impact: Short

The Snowflake AI Research Team has launched Snowflake Arctic, an LLM that they say offers high-quality performance at a lower cost, especially in tasks like SQL generation, coding, and instruction-following benchmarks. The model is now available on platforms including Hugging Face, Snowflake Cortex, AWS, Microsoft Azure, and the NVIDIA API catalog. They stated that "Arctic uses a unique Dense-MoE Hybrid transformer architecture. It combines a 10B dense transformer model with a residual 128×3.66B MoE MLP resulting in 480B total and 17B active parameters chosen using a top-2 gating. It was designed and trained using the following three key insights and innovations." [Snowflake Arctic - LLM for Enterprise AI] Explore more of our coverage of: Snowflake AI, LLM Performance, Dense-MoE Hybrid Transformer. Share this story by email

Hugging Face Launches Imgsys, Image Evaluation Arena

First impacted: AI developers, Graphic designers
Time to impact: Short

Hugging Face has launched Imgsys, an image evaluation arena. RealVisXL V4.0 and Proteus are currently leading the competition. [imgsys.org - a Hugging Face Space by fal-ai] Explore more of our coverage of: Hugging Face Spaces, Imgsys Chatbot, AI Image Generation. Share this story by email

Apple Launches OpenELM Language Model

[Paper page - OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework] Explore more of our coverage of: Open-source Language Models, Apple OpenELM, AI Efficiency. Share this story by email

MambaByte Video Proposes Efficiency in AI Training

[MambaByte: Token-Free Language Modeling] Explore more of our coverage of: AI Training, Language Modeling, Token-Level Speculation. Share this story by email

OpenAI Launches Instruction Hierarchy Initiative to Enhance LLM Robustness Against Deceptive Actions

[Careers at OpenAI] Explore more of our coverage of: OpenAI, Instruction Hierarchy, LLM Resilience. Share this story by email