Ollama supports a list of models available on ollama.com/library
Here are some example models that can be downloaded:
| Model | Parameters | Size | Download |
|---|---|---|---|
| Llama 3.2 | 3B | 2GB | ollama run llama3.2 |
| Llama 3.1 | 8B | 4.7GB | ollama run llama3.1 |
| Llama 3.1 | 70B | 40GB | ollama run llama3.1:70b |
| Llama 3.1 | 405B | 231GB | ollama run llama3.1:405b |
| Phi 3 Mini | 3.8B | 2.3GB | ollama run phi3 |
| Phi 3 Medium | 14B | 7.9GB | ollama run phi3:medium |
| Gemma 2 | 2B | 1.6GB | ollama run gemma2:2b |
| Gemma 2 | 9B | 5.5GB | ollama run gemma2 |
| Gemma 2 | 27B | 16GB | ollama run gemma2:27b |
| Mistral | 7B | 4.1GB | ollama run mistral |
| mistral-small | 22b | 13GB | ollama run mistral-small |
| Moondream 2 | 1.4B | 829MB | ollama run moondream |
| Neural Chat | 7B | 4.1GB | ollama run neural-chat |
| Starling | 7B | 4.1GB | ollama run starling-lm |
| Code Llama | 7B | 3.8GB | ollama run codellama |
| Llama 2 Uncensored | 7B | 3.8GB | ollama run llama2-uncensored |
| LLaVA | 7B | 4.5GB | ollama run llava |
| Solar | 10.7B | 6.1GB | ollama run solar |
Note:
You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
Supported Models
Ollama supports a wide range of Large Language Models (LLMs) tailored for various tasks, such as coding, general-purpose language understanding, multilingual applications, and scientific discussions. These models vary in size and capabilities, with parameter counts ranging from a few billion to over 400 billion. Below is a table listing the supported models along with a brief description and associated tags.
| Model Name | Description | Tags |
|---|---|---|
| Lama 3.2 | Meta’s Llama 3.2 goes small with 1B and 3B models. | Tools |
| Llama 3.1 | Meta’s latest state-of-the-art LLM available in 8B, 70B, and 405B parameter sizes. | Tools |
| Gemma 2 | High-performing model by Google, available in 2B, 9B, and 27B parameter sizes. | – |
| Mistral-Nemo | A state-of-the-art 12B model with 128k context length, developed by Mistral AI and NVIDIA. | Tools |
| Mistral Large 2 | Mistral’s flagship model excels in code generation, mathematics, and reasoning, supporting a 128k context window. | Tools |
| Qwen 2 | A series of large language models by Alibaba, ranging from 0.5B to 72B parameters. | Tools |
| Qwen2.5 | Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, a range of base language models and instruction-tuned models are released, with sizes ranging from 0.5 to 72 billion parameters. | Tools |
| DeepSeek Coder V2 | Open-source Mixture-of-Experts code language model comparable to GPT-4 Turbo in coding tasks. | Code |
| Phi-3 | Lightweight LLMs by Microsoft in 3B (Mini) and 14B (Medium) parameter sizes. | – |
| Mistral | The updated 7B model by Mistral AI. | Tools |
| Mixtral | Mixture of Experts model with open weights available in 8x7B and 8x22B parameter sizes. | Tools |
| mistral-small | Mistral Small is a lightweight model designed for cost-effective use in tasks like translation and summarization. | Tools |
| CodeGemma | Powerful models for coding tasks like code generation and natural language understanding. | Code |
| Command R | LLM optimized for conversational interaction and long context tasks. | Tools |
| Command R+ | Scalable LLM designed for enterprise use cases. | Tools |
| LLaVA | Multimodal model combining vision encoder and Vicuna for visual and language understanding, updated to version 1.6. | Vision |
| Llama 3 | Meta’s most capable openly available LLM. | – |
| Gemma | Lightweight state-of-the-art models by Google DeepMind, updated to version 1.1. | – |
| Qwen | Large language models by Alibaba Cloud, available in sizes up to 110B parameters. | – |
| Llama 2 | A collection of foundation language models ranging from 7B to 70B parameters. | – |
| CodeLlama | LLM that generates and discusses code using text prompts, with sizes up to 70B parameters. | Code |
| Nomic Embed Text | High-performing open embedding model with a large token context window. | Embedding |
| Dolphin Mixtral | Uncensored models based on Mixtral, optimized for coding tasks, available in 8x7B and 8x22B sizes. | – |
| Phi | A 2.7B language model by Microsoft Research, excelling in reasoning and language understanding. | – |
| Llama 2 Uncensored | Uncensored version of Llama 2 by George Sung and Jarrad Hope. | – |
| DeepSeek Coder | Capable coding model trained on two trillion code and natural language tokens. | Code |
| MXBAI Embed Large | State-of-the-art large embedding model by Mixedbread.ai. | Embedding |
| Dolphin Mistral | Uncensored model based on Mistral, excelling at coding tasks, updated to version 2.8. | – |
| Zephyr | Fine-tuned versions of Mistral and Mixtral models designed to act as helpful assistants. | – |
| StarCoder 2 | Next-generation open code LLMs available in 3B, 7B, and 15B parameter sizes. | Code |
| Dolphin Llama 3 | A new model by Eric Hartford based on Llama 3, available in 8B and 70B sizes. | – |
| Orca Mini | General-purpose model ranging from 3B to 70B parameters, suitable for entry-level hardware. | – |
| Yi 1.5 | High-performing bilingual language model. | – |
| Mistral OpenOrca | Fine-tuned on the OpenOrca dataset, based on the Mistral 7B model. | – |
| LLaVA Llama 3 | LLaVA model fine-tuned from Llama 3 Instruct, with better benchmark scores. | Vision |
| StarCoder | Code generation model trained on over 80 programming languages. | Code |
| Vicuna | General-use chat model based on Llama and Llama 2, with context sizes from 2K to 16K. | – |
| TinyLlama | An open project to train a compact 1.1B Llama model on 3 trillion tokens. | – |
| Llama 2 Chinese | Llama 2 model fine-tuned for improved Chinese dialogue ability. | – |
| Codestral | Mistral AI’s first-ever code model designed for code generation tasks. | Code |
| Wizard Vicuna | Uncensored model based on Llama 2, available in 7B, 13B, and 30B parameter sizes. | – |
| Nous Hermes 2 | Powerful models by Nous Research, excelling in scientific discussion and coding tasks. | – |
| CodeGeeX 4 | Versatile model for AI software development scenarios, including code completion. | Code |
| OpenChat | Open-source models trained on diverse data, surpassing ChatGPT on various benchmarks. | – |
| Aya 23 | A state-of-the-art multilingual model supporting 23 languages, released by Cohere. | – |
| Granite Code | IBM’s open foundation models for Code Intelligence. | Code |
| WizardLM 2 | Advanced LLM by Microsoft AI, optimized for complex chat, multilingual, reasoning, and agent use cases. | – |
| TinyDolphin | An experimental 1.1B parameter model based on TinyLlama and trained on the Dolphin 2.8 dataset by Eric Hartford. | – |
| CodeQwen | A large language model pre-trained on extensive code data. | Code |
| WizardCoder | A state-of-the-art model for code generation. | Code |
| Stable Code | A coding model with instruct and code completion variants, competitive with larger models like CodeLlama 7B. | Code |
| All MiniLM | Embedding models trained on very large sentence-level datasets. | Embedding |
| OpenHermes | A 7B model fine-tuned by Teknium on Mistral using fully open datasets. | – |
| Stable LM 2 | A multilingual model trained on English, Spanish, German, and other languages, available in 1.6B and 12B sizes. | – |
| Wizard Math | Model focused on solving math and logic problems. | – |