#Nvidia lance #Nemotron70B, #IA qui explose #GPT4o
👉Séisme dans l’industrie de l’IA
👉Un modèle qui surpasse ceux des leaders comme OpenAI et Anthropic
👉Nvidia a affiné le modèle open-source #Llama3_1 avec des techniques d’entraînement très avancées
www.lebigdata.fr/nvidia-lance...

Meta lanzó Llama 3.1, su modelo más potente (405B), ahora con una ventana de contexto masiva de 128K tokens (16x más que Llama 3). Es el modelo abierto más grande, competitivo con GPT-4. Soporta 8 idiomas y está diseñado para long-form y código. #Meta #Llama3_1 #OpenAI #LLM #GadgetsTimesRD

So any model trained/fine-tuned with llama outputs — even if it's not based on llama's architecture at all — must be prefixed with "llama" 😲 .

e.g. huggingface.co/ai4privacy/l... (ModernBert fine-tune)

License: www.llama.com/llama3_1/lic...

Gemma‑2‑2B‑Instruct teams up with Llama‑3.1‑8B‑Instruct and slashes 99.9 tokens on a 248‑prompt benchmark. Curious how they got such a boost? Dive into the details. #Gemma2B #Llama3_1 #TokenSavings

🔗 aidailypost.com/news/gemma-2...

AMD just set up a Llama 3.1 8B pre‑training benchmark for MLPerf using random weights. Curious how a blank‑slate model stacks up? Dive into the details! #AMD #Llama3_1 #MLPerf

🔗 aidailypost.com/news/amd-bui...

Just saw NVIDIA’s NVFP4 recipe slash training time and costs on Blackwell Ultra GPUs—MLPerf scores are soaring and Llama 3.1 cranks out faster than ever. Want the nitty‑gritty on how GPU acceleration is reshaping LLM training? Dive in! #NVFP4 #MLPerf #Llama3_1

🔗 aidailypost.com/news/nvidias...