IBM's latest Granite 3.1 models and their new embedding models are now available via Ollama. Dense models are designed for tool-based use cases. ollama run granite3.1-dense:2b ollama run granite3.1-dense:8b Mixture of expert models are designed for low latency usage. ollama run granite3-moe:1b ollama run granite3-moe:3b Embedding models are smaller in size in comparison to others with similar performance. English only: ollama pull granite-embedding:30m Multilingual: ollama pull granite-embedding:278m Ollama model pages: Dense: https://lnkd.in/g4GkKk9P MoE: https://lnkd.in/gw84-i-e Embedding: https://lnkd.in/gNCvrB-Q
Thinking of using these with Continue.dev. Think I can combine all these and make it work?
IBM's Granite 3.1 models showcase versatility with dense, MoE, and embedding options!
I am interested to see how this generation perform. The last struggled against llama 3.2 3b for tool use. I managed to get it to only execute tools about 50% of the time vs 90% for llama3.2
Tried previous versions and was disappointed a little. Will check these latest models. Hope they made some updates
How may the embedding dimension of the emmbeding model?
Let’s try it 🙌
Can't wait
Get started with Granite 3.1 today ⚡