Ollama Model Gallery

What this article covers

• Overview of popular Ollama models.
• Organized by use case.
• Model sizes, quantization options, and hardware requirements.
• Tips for quick selection.

Introduction: Ollama Model Gallery

Ollama provides access to a growing library of local language and multimodal models. Chat models like Llama and Qwen, coding models like Qwen Coder, vision models like Llava, and embedding models like nomic-embed-text are just a few examples. The selection expands rapidly, and multiple options exist for nearly every use case.

This article assembles a gallery of the most important Ollama models.

Key terminology

• 7B, 13B, 70B: Model size in billions of parameters.
• Instruct: Trained on instructions.
• Chat: Optimized for dialogue.
• Coder: Optimized for programming.
• Vision: Image processing.
• Embedding: Vector representation.
• Quantization: Reduced model precision.
• GGUF: Container format for models.

Chat models

Model	Size	Highlights
llama3.1	8B, 70B	Solid all-rounder
qwen2.5	7B, 14B, 32B	Strong multilingual support
mistral	7B	Fast, compact
gemma2	9B, 27B	Google model
phi4	14B	Strong reasoning

Coding models

Model	Size	Use case
qwen2.5-coder	1.5B-32B	Code completion
codellama	7B-70B	Meta code model
deepseek-coder	1.3B-33B	Versatile coder

Vision models

Model	Size	Use case
llava	7B, 13B	Image description
llava-llama3	8B	Compact image analysis
bakllava	7B	Better vision quality
moondream	2B	Very small, fast

Reasoning models

Model	Size	Use case
deepseek-r1	1.5B-70B	Logical reasoning
qwen2.5	7B-72B	Math and logic
phi4	14B	Smaller reasoning

Embedding models

Model	Dimension	Use case
nomic-embed-text	768	Standard embeddings
mxbai-embed-large	1024	High quality
snowflake-arctic-embed	768/1024	Document retrieval
all-minilm	384	Very small

Specialized models

Model	Use case
nomic-embed-text-v1.5	Improved embeddings
granite3.2-vision	IBM vision
command-r	Cohere instruct
aya	Multilingual

Download a model

ollama pull llama3.1
ollama pull qwen2.5-coder:14b
ollama pull llava
ollama pull nomic-embed-text

Choose a model

Use case	Model
General chat	llama3.1, qwen2.5
Coding	qwen2.5-coder, deepseek-coder
Vision	llava, bakllava, moondream
Reasoning	deepseek-r1, qwen2.5
RAG	nomic-embed-text, mxbai-embed-large
German	qwen2.5, aya

Hardware requirements

Model size	VRAM Q4	RAM CPU
3B	2-4 GB	6-8 GB
7B	4-8 GB	12-16 GB
13B	8-12 GB	20-32 GB
30B	18-24 GB	40-64 GB
70B	40-48 GB	96-128 GB

Tips

• Start with 7B models.
• Pay attention to quantization options.
• Use different models for chat and RAG.
• Test against your own requirements.
• Compare models on Ollama.com.
• Prefer updated versions.

Further reading and resources

• BotServ.de Ollama Model Recommendations
• BotServ.de Ollama Model Benchmarks
• BotServ.de Ollama Embedding Comparison
• BotServ.de Ollama Vision API

FAQ: Ollama Model Gallery

Which model is best for beginners? llama3.1:8b or qwen2.5:7b.

Are there German models? Yes, Qwen 2.5 and Aya have good multilingual support.

What’s the smallest vision model? moondream:2b.

Which model for coding? qwen2.5-coder:14b or deepseek-coder.

Which models for RAG? A solid chat model plus an embedding model.

Sources and further reading

• Ollama Library: https://ollama.com/library
• Hugging Face: https://huggingface.co/
• LMSYS Chatbot Arena: https://chat.lmsys.org/

Summary: Ollama Model Gallery

The Ollama library includes many models for chat, coding, vision, reasoning, and embeddings. Llama, Qwen, and Mistral are solid all-rounders; Qwen Coder and DeepSeek Coder excel at programming; Llava and Moondream handle images; nomic-embed-text powers semantic search. The right choice depends on your hardware, use case, and desired quality. Starting with a typical 7B variant usually works well.