Ollama Models Hub — Popular Local LLMs for Windows (Llama 3, Mistral, Qwen 2.5)

Copy any ollama pull command below and paste it into Command Prompt or PowerShell. The model downloads once and then runs fully offline.

Llama 3 — Meta

Llama 3 is Meta's flagship open-source model. It delivers an excellent balance of quality, speed and hardware compatibility, making it the best starting point for most Windows users.

Variant	Size	VRAM needed	Pull command
8B (default)	4.7 GB	6 GB+	`ollama pull llama3`
8B Q4_K_M	4.4 GB	5 GB+	`ollama pull llama3:8b-instruct-q4_K_M`
70B	40 GB	48 GB+	`ollama pull llama3:70b`

cmd.exe

C:\> ollama pull llama3

C:\> ollama run llama3

Full guide: Llama 3 on Windows

Fast & efficient

Mistral — Mistral AI

Mistral 7B is one of the fastest models per token on consumer hardware. It excels at coding, summarisation and quick Q&A. Runs well even on GPUs with 6 GB VRAM or CPU-only.

Variant	Size	VRAM needed	Pull command
7B (default)	4.1 GB	5 GB+	`ollama pull mistral`
Mistral Nemo 12B	7.1 GB	8 GB+	`ollama pull mistral-nemo`

cmd.exe

C:\> ollama pull mistral

C:\> ollama run mistral

Full guide: Mistral on Windows

Multilingual & reasoning

Qwen 2.5 — Alibaba

Qwen 2.5 leads open-source benchmarks in reasoning, math and multilingual tasks. Available in sizes from 0.5B to 72B. Strong choice for non-English languages and technical work.

Variant	Size	VRAM needed	Pull command
7B	4.4 GB	6 GB+	`ollama pull qwen2.5`
14B	8.7 GB	10 GB+	`ollama pull qwen2.5:14b`
Coder 7B	4.4 GB	6 GB+	`ollama pull qwen2.5-coder`

cmd.exe

C:\> ollama pull qwen2.5

C:\> ollama run qwen2.5

Full guide: Qwen 2.5 on Windows

Compact & capable

Gemma 2 — Google

Gemma 2 from Google punches above its weight for its size. The 2B variant runs on almost any hardware including CPU-only. Good for embedded use cases and machines with limited RAM.

Variant	Size	VRAM needed	Pull command
2B	1.6 GB	3 GB+	`ollama pull gemma2:2b`
9B	5.4 GB	7 GB+	`ollama pull gemma2`
27B	16 GB	20 GB+	`ollama pull gemma2:27b`

cmd.exe

C:\> ollama pull gemma2:2b

C:\> ollama run gemma2:2b

Small & fast

Phi-4 — Microsoft

Phi-4 is Microsoft's small language model optimised for instruction following and coding. At 14B parameters it delivers quality comparable to much larger models on many benchmarks, while fitting in 8 GB VRAM.