Releasing METAGENE-1: In collaboration with researchers from USC, we're open-sourcing a state-of-the-art 7B parameter Metagenomic Foundation Model. Enabling planetary-scale pathogen detection and reducing the risk of pandemics in the age of exponential biology.

MiniMax, the company behind Hailuo AI (MiniMax), has joined the open source community by releasing both models and demos of MiniMax-Text-01 & MiniMax-VL-01🔥 huggingface.co/MiniMaxAI/Mini… huggingface.co/MiniMaxAI/Mini… ✨ MiniMax-text-01: - 456B with 45.9B activated per token - Combines Lightning

InternLM3-8B🔥 Trained on just 4T tokens, it outperforms Llama3.1-8B and Qwen2.5-7B in reasoning tasks, at 75% lower cost! huggingface.co/collections/in…

BIG release by DeepSeek 🔥🔥🔥 DeepSeek-R1 & DeepSeek-R1-Zero: two 660B reasoning models are here, alongside 6 distilled dense models (based on Llama & Qwen) for the community! huggingface.co/deepseek-ai huggingface.co/deepseek-ai/De… ✨ MIT License : enabling distillation for custom

What happened yesterday in the Chinese AI community? 🚀 T2A-01-HD 👉 hailuo.ai/audio MiniMax's Text-to-Audio model, now in Hailuo AI, offers 300+ voices in 17+ languages and instant emotional voice cloning. Tare 👉 trae.ai A new coding tool by Bytedance

OpenAI Operator looks great here is a open source version of it you can use now see an agent that uses your browser to perform tasks for you developers get started in a few lines of code pip install 'ai-gradio[browser]' import gradio as gr import ai_gradio gr.load(

I don’t get the “is scale all you need?” debate. Here’s how I see it: accuracy = resources * (accuracy / resources), where - resources is how much you’ve scaled up data|compute, and - accuracy / resources is the data|compute efficiency of your method.

Qwen2.5-1M 🔥the long-context version of Qwen2.5, supports up to 1M tokens by Qwen huggingface.co/collections/Qw…

So many exciting releases coming from the Chinese community this month 🔥 huggingface.co/collections/zh… LLMs: ✨ Qwen2.5 -1M by Qwen huggingface.co/collections/Qw… ✨ InternLM3-8B-Instruct by InternLM huggingface.co/internlm/inter… ✨ MiniMax-Text-01 by MiniMax AI

On the last day before the Spring Festival holiday in China, DeepSeek released a NEW work on Hugging Face 🤯 Janus-Pro🔥 autoregressive framework that unifies multimodal understanding and generation huggingface.co/deepseek-ai/Ja… ✨ 1B / 7B ✨ MIT License

YuE 乐🎵 7B open music foundation models released by the community M-A-P huggingface.co/m-a-p/YuE-s1-7… ✨ Transforms lyrics into full songs ✨ Support English/Chinese/Japanese/Korean

New MoE model 🔥 Qwen2.5-Max 🧧 New Year's gift from Qwen huggingface.co/spaces/Qwen/Qw…

The US Copyright office has ruled that AI/human combined work can be copyrighted as long as a human is adding, changing or selecting elements. Prompts alone do not usually produce copyrighted work. Everything is case-by-case, but the report is clear and thoughtful about reasoning

Latest LLM integrations show a shift in agent architecture: Instead of single-model inference, new systems use specialized models for different tasks, improving accuracy and reducing latency.

GRPO actually works well for optimizing LLMs if you avoid reward hacking. Key is tuning 3 things: system prompt (1st person), learning rate, and KL coefficient. Simple but effective for model improvement.

📢 Terminology Correction: DeepSeek-R1’s code and models are released under the MIT License.

Jan Leike I don’t want to provide my world-class expertise just for you to hoard crowd-sourced prompts and construct elaborate security theater performances to appease investors who are foolish enough to believe guardrails=safety. I’m allergic to money, so don’t bother. My incentives are

Most people think success is about having more. More funding, more headcount, more connections. They’re wrong. The best don’t win because they have more. They win because they do more with whatever they have. Constraints don’t slow them down. They sharpen their thinking. They

Paperback? $1 per 10,000 tokens Customer support? $1 per 100 tokens SWE/doctor/lawyer? $1 per token But LLMs give you a million tokens to the dollar (and falling) – Noam Shazeer

Manus is IMPRESSIVE🚀 Just 3 hours after its release, the open-source community responded with: 🔓 Open Manus github.com/mannaandpoem/O… 🦉 Owl by CAMEL github.com/camel-ai/owl The community moves FAST⚡