AI: Encompasses all techniques and technologies aimed at simulating human intelligence.
ML: A subset of AI focused on creating models that learn from data.
DL: A subset of ML that uses deep neural networks to handle complex and high-dimensional data.
Generative AI: Focuses on creating new content and data, using advanced models to generate outputs that mimic or extend the training data.
foundation model
IBM foundation model
Meta foundation model
Google foundation model
Code Llama foundation model
Mistral AI foundation model
BigScience Mistral AI foundation model
FACTS is a novel benchmark from Google DeepMind and Google Research designed to evaluate the factual accuracy and grounding of AI models.
The FACTS Grounding benchmark evaluates the ability of Large Language Models (LLMs) to generate factually accurate responses grounded in provided long-form documents, encompassing a variety of domains. FACTS Grounding moves beyond simple factual question-answering by assessing whether LLM responses are fully grounded to the provided context and correctly synthesize information from a long context document. By providing a standardized evaluation framework, FACTS Grounding aims to promote the development of LLMs that are both knowledgeable and trustworthy, facilitating their responsible deployment in real-world applications.
Publisher: Google
Based on Gemma 2 and SigLIP open components, PaliGemma 2 is a vision-language model with superior fine-tuning performance. It excels at multilingual image-text tasks like captioning, object detection, and visual QA.
google/paligemma-2
The PaliGemma family of models is inspired by PaLI-3 and based on open components such as the SigLIP vision model and Gemma 2 language models.
Publisher: Meta
Llama 3.3 is a 70B LLM optimized for multilingual dialogue. It leverages SFT and RLHF tuning to ensure exceptional performance, safety, and helpfulness.
metaresearch/llama-3.3
The Meta Llama 3.3 multilingual large language model (LLM) is an instruction tuned generative model in 70B (text in/text out).
Publisher: Microsoft
A lightweight, open model family performant in reasoning and instruction adherence, Phi-3.5 has a 128K token context and advanced safety optimizations. Variants: Mini-instruct, MoE-instruct, and Vision-instruct.
Microsoft/phi-3
Tiny but Mighty: The Phi-3 small language models with big potential.
Publisher: Qwen / Alibaba
QwQ-32B-Preview, a 32.5B parameter causal language model featuring 64 layers and a 32K token context. An experimental model designed for advanced reasoning, it’s ideal for math, coding, and analytical tasks.
qwen-lm/qwq-32b-preview
QwQ-32B-Preview is an experimental research model developed by the Qwen Team, focused on advancing AI reasoning capabilities.
Flash Publisher: Google
Gemini 2.0 Flash is an experimental release with faster TTFT, enhanced multimodal understanding, native image generation, and controllable text-to-speech. It powers real-time vision and audio streaming applications.
google/gemini-2.0-flash-api
A new family of multimodal models from Google DeepMind