the-decoder4d
ModernBERT is a "workhorse model" that brings faster, cheaper text processing for tasks like RAG
Answer.AI and LightOn announced ModernBERT, a new open-source language model that improves upon Google's BERT in speed, efficiency, and quality. The encoder-only model processes text up to four times ...
the-decoder7d
Google's Gemini 2.0 "Flash Thinking" is Google's answer to OpenAI's o1
Google has unveiled an experimental version of Gemini 2.0 that shows its step-by-step reasoning process. The model performs well on early human-rated benchmarks. Google has released an experimental ...
the-decoder5d
OpenAI unveils o3, its most advanced reasoning model yet
OpenAI has announced o3, a new AI model that achieves breakthrough performance in complex reasoning tasks. A cost-effective mini version is set to launch in late January 2025, followed by the full ...
the-decoder5d
Anthropic's Claude AI cooperates better than OpenAI and Google models, study finds
A new research paper reveals significant differences in how AI language models work together, with Anthropic's Claude 3.5 Sonnet showing superior cooperation skills compared to competitors. The ...
the-decoder5d
Meta plans multiple Llama 4 releases in 2025, focusing on reasoning and speech
Betting big on reasoning and AI agents For 2025, Meta plans to accelerate Llama's development with several Llama 4 releases. According to Zuckerberg, training Llama 4 could require nearly ten times ...
the-decoder7d
Falcon 3 series sets new benchmarks for open-source LLMs on a single GPU
Abu Dhabi's Technology Innovation Institute (TII) has released Falcon 3, its latest generation of open source AI models. According to TII, the new models set performance records specifically for AI ...
the-decoder5d
OpenAI's o3 model shows major gains through reinforcement learning scaling
OpenAI reports breakthrough performance with its new o3 reasoning model. The company attributes much of this progress to reinforcement learning, a well-established AI training method that offers ...
the-decoder8d
Study shows: 'Test-time compute scaling' is a path to better AI systems
Researchers at Hugging Face have demonstrated significant performance improvements in open-source language models by intelligently scaling compute during inference, drawing inspiration from OpenAI's ...
the-decoder6d
ElevenLabs launches Flash, its fastest text-to-speech AI yet
ElevenLabs has unveiled Flash, its newest text-to-speech model built specifically for speed. The system can transform text into speech in just 75 milliseconds - not including network and application ...
the-decoder5d
AI models can only pretend to follow human rules, Anthropic study finds
A new study by Anthropic and Redwood Research shows that large language models like Claude can pretend to follow safety guidelines while pursuing different objectives behind the scenes. For the first ...
the-decoder10d
Google launches Whisk, an AI tool that combines multiple images for generation
Google Labs has released Whisk, its latest generative AI experiment, in the United States. Unlike traditional image generators, which mostly rely on text prompts, Whisk focuses on using images as the ...
the-decoder8d
OpenAI makes ChatGPT available through traditional phone calls and WhatsApp
OpenAI is making ChatGPT more accessible by letting users interact with the AI assistant through phone calls and WhatsApp messages. U.S. users can now call ChatGPT for free at 1-800-242-8478, with a ...