Nvidia just dropped their Nemotron 70b model, and it’s blowing past competitors like OpenAI’s GPT-4o and Anthropic’s Claude-3. It’s completely open-source and free, making it a game changer for developers and researchers. If you’ve been keeping an eye on the AI landscape, this is a pivotal moment.
The Llama-3.1-Nemotron-70B-Instruct model comes with 70 billion parameters, making it incredibly powerful when it comes to reasoning and handling complex queries.
But while Nemotron is great, it has issues with the strawberry test. When playing with Hugging chat and the chat in Nvidia’s Nemotron NIM page, the model displayed the classic “There’s only 2 r’s in strawberry”. But it corrected itself after I stepped it out.
When run on Ollama, it nails the strawberry test on the first try, showing there’s likely some differences across platforms.
Here’s how to try it out:
- Hugging Chat: Head to huggingface.co/chat and select the
nvidia/Llama-3.1-Nemotron-70B-Instruct-HFmodel. - Ollama:
ollama run nemotron. Need an supported graphics card, and the model is 42 Gig
https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
This is a huge step forward for AI, and with Nemotron being free and open-source, it’s a must-try. But it’ll be outdated in a couple days, I’m sure.
0 Comments