What large language models does Ollama support and how to choose the right model?

Ollama supports various open-source Large Language Models, mainly including:

1. Meta Llama Series:

• llama2 - Llama 2 7B/13B/70B
• llama3 - Llama 3 8B/70B
• llama3.1 - Llama 3.1 8B/70B/405B
• llama3.2 - Llama 3.2 1B/3B

2. Mistral AI Series:

• mistral - Mistral 7B
• mixtral - Mixtral 8x7B
• mixtral:8x22b - Mixtral 8x22B

3. Google Gemma Series:

• gemma - Gemma 2B/7B
• gemma2 - Gemma 2 9B/27B

4. Code-Specific Models:

• codellama - Code Llama 7B/13B/34B
• deepseek-coder - DeepSeek Coder

5. Other Popular Models:

• qwen - Qwen series
• phi - Microsoft Phi series
• gemma2:9b - Lightweight model
• tinyllama - TinyLlama 1.1B

Model Selection Recommendations:

1. General Conversation: Recommend llama3.1:8b or mistral:7b
2. Code Generation: Recommend codellama:13b or deepseek-coder
3. Lightweight Applications: Recommend phi:3.8b or gemma2:9b
4. High-Quality Output: Recommend llama3.1:70b (requires more memory)

Model Version Specification:

bash
ollama run llama3.1:8b
ollama run mistral:7b-instruct
ollama run codellama:13b-python

View Available Models: Visit https://ollama.com/library to see all available models and their variants.