Ollama quick reference

Ollama commands by GPU fit

Use this page when you already know you want Ollama and need a command that matches your local hardware. Start with green GPU badges; yellow means RAM/offload tradeoffs.

Open interactive calculator

Commands34
8GB clean fits8
12GB clean fits14
24GB clean fits26
Install

Install Ollama first, then use one of the commands below.

Pick green

Green badges are the clean local starting point for that GPU class.

Check context

Long context, tools, and parallel runs can increase real memory use.

8GB to 12GB starter commands

Good first commands for small local chat, simple agents, and quick tests
9 commands

Qwen3 4B Thinking 2507

4.02B Q4 about 3.2 GB Alibaba CodingAgentsReasoning
ollama run qwen3:4b-thinking-2507-q4_K_M
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Small local reasoning, routing, tool decisions, and lightweight coding on 6GB-8GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

Gemma 3 4B

4B Q4 about 3.5 GB Google AgentsVisionChat
ollama run gemma3:4b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Small multimodal local assistant and low-resource setups

Open model page -> Calculator keeps hardware-specific checks.

Qwen2.5 Coder 7B

7B Q4 about 5.5 GB Alibaba CodingAgentsChat
ollama run qwen2.5-coder:7b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Small local coding assistant and agent tool generation

Open model page -> Calculator keeps hardware-specific checks.

Mistral 7B

7B Q4 about 5.5 GB Mistral AI AgentsChat
ollama run mistral:7b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Fast local chat and simple agent tasks

Open model page -> Calculator keeps hardware-specific checks.

Llama 3.1 8B Instruct

8B Q4 about 6 GB Meta AgentsChat
ollama run llama3.1:8b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Fast local chat, lightweight agents, low-cost local testing

Open model page -> Calculator keeps hardware-specific checks.

Qwen3 8B

8B Q4 about 6 GB Alibaba CodingAgentsReasoning
ollama run qwen3:8b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Fast general local assistant with reasoning/coding balance

Open model page -> Calculator keeps hardware-specific checks.

DeepSeek-R1-0528-Qwen3-8B

8B Q4 about 6 GB DeepSeek CodingReasoningChat
ollama run deepseek-r1:8b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Updated local reasoning experiments, coding logic checks, and step-by-step technical analysis

Open model page -> Calculator keeps hardware-specific checks.

Qwen3-VL 8B Instruct

8B Q4 about 6.5 GB Alibaba AgentsReasoningVision
ollama run qwen3-vl:8b
8GB: Runs locally12GB: Runs locally24GB: Runs locally

Local image understanding, OCR-style document triage, screenshots, and lightweight multimodal agent routing

Open model page -> Calculator keeps hardware-specific checks.

Qwen3.5 9B

9B Q4 about 6.6 GB Alibaba CodingAgentsReasoning
ollama run qwen3.5:9b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Modern multimodal local assistant, agent experiments, and coding support on mainstream GPUs

Open model page -> Calculator keeps hardware-specific checks.

12GB to 24GB workstation commands

Useful for coding assistants, local agents, and stronger general-purpose models
7 commands

Gemma 3 12B

12B Q4 about 9 GB Google AgentsReasoningVision
ollama run gemma3:12b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Balanced multimodal local chat on 12GB+ GPUs

Open model page -> Calculator keeps hardware-specific checks.

Gemma 4 E4B

4B Q4 about 9.6 GB Google CodingAgentsReasoning
ollama run gemma4:e4b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Efficient multimodal local assistant and edge-style agent workflows

Open model page -> Calculator keeps hardware-specific checks.

Qwen2.5 Coder 14B

14B Q4 about 10.5 GB Alibaba CodingAgentsReasoning
ollama run qwen2.5-coder:14b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Local coding, scripts, repo assistance, technical agents

Open model page -> Calculator keeps hardware-specific checks.

DeepSeek R1 Distill Qwen 14B

14B Q4 about 10.5 GB DeepSeek CodingReasoningChat
ollama run deepseek-r1:14b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Local reasoning and debugging on 12GB/16GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

Phi-4 14B

14B Q4 about 10.5 GB Microsoft CodingAgentsReasoning
ollama run phi4:14b
8GB: RAM offload12GB: Runs locally24GB: Runs locally

Compact reasoning and technical assistant on 12GB/16GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

gpt-oss 20B

20B Q4 about 14 GB OpenAI CodingAgentsReasoning
ollama run gpt-oss:20b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Local reasoning, agent planning, and tool-use workflows on 16GB+ GPUs

Open model page -> Calculator keeps hardware-specific checks.

Devstral Small 2 24B

24B Q4 about 15 GB Mistral AI CodingAgentsReasoning
ollama run devstral-small-2
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Software engineering agents, repo navigation, patch planning, and local coding workflows

Open model page -> Calculator keeps hardware-specific checks.

24GB workstation commands

Use these when you have RTX 3090/4090-class headroom or similar
11 commands

Qwen3.5 27B

27B Q4 about 17 GB Alibaba CodingAgentsReasoning
ollama run qwen3.5:27b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

24GB-class multimodal agent, coding assistant, and reasoning workloads

Open model page -> Calculator keeps hardware-specific checks.

Qwen3.6 27B

27B Q4 about 17 GB Alibaba CodingAgentsReasoning
ollama run qwen3.6:27b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Newest 27B-class local multimodal coding, agent, and reasoning workloads on 24GB-class GPUs

Open model page -> Calculator keeps hardware-specific checks.

Gemma 3 27B

27B Q4 about 18 GB Google AgentsReasoningVision
ollama run gemma3:27b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

High-quality multimodal local assistant on 24GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

Qwen3 30B-A3B Instruct 2507

30.5B Q4 about 19 GB Alibaba CodingAgentsReasoning
ollama run qwen3:30b-a3b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

General local reasoning, tool agents, multilingual writing, and coding on 24GB+ GPUs

Open model page -> Calculator keeps hardware-specific checks.

Qwen3-Coder 30B-A3B

30B Q4 about 19 GB Alibaba CodingAgentsReasoning
ollama run qwen3-coder:30b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Agentic coding, repository-scale local code review, and tool-heavy development loops

Open model page -> Calculator keeps hardware-specific checks.

GLM-4.7-Flash

30B Q4 about 19 GB Z.ai CodingAgentsReasoning
ollama run glm-4.7-flash:q4_K_M
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Local coding agents, terminal workflows, tool-heavy engineering tasks, and 30B-class reasoning on 24GB+ GPUs

Open model page -> Calculator keeps hardware-specific checks.

Gemma 4 31B

31B Q4 about 20 GB Google CodingAgentsReasoning
ollama run gemma4:31b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

High-quality multimodal reasoning, coding assistants, and local-first agent workflows

Open model page -> Calculator keeps hardware-specific checks.

Qwen3-VL 30B-A3B Instruct

30B Q4 about 20 GB Alibaba AgentsReasoningVision
ollama run qwen3-vl:30b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Higher-quality local multimodal reasoning, screenshot analysis, document/image extraction, and GUI-agent planning

Open model page -> Calculator keeps hardware-specific checks.

Qwen2.5 Coder 32B

32B Q4 about 21 GB Alibaba CodingAgentsReasoning
ollama run qwen2.5-coder:32b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Strong local coding and architecture work on 24GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

DeepSeek R1 Distill Qwen 32B

32B Q4 about 21 GB DeepSeek CodingReasoningChat
ollama run deepseek-r1:32b
8GB: RAM offload12GB: RAM offload24GB: Runs locally

Heavy local reasoning on 24GB GPUs

Open model page -> Calculator keeps hardware-specific checks.

Qwen3.6 35B-A3B

35B Q4 about 24 GB Alibaba CodingAgentsReasoning
ollama run qwen3.6:35b-a3b
8GB: RAM offload12GB: RAM offload24GB: RAM offload

Agentic coding and multimodal reasoning when 27B is not enough and 32GB-class headroom is available

Open model page -> Calculator keeps hardware-specific checks.

Large or fallback-first commands

Check hardware carefully before running these locally
7 commands

Mixtral 8x7B

46.7B Q4 about 28 GB Mistral AI CodingAgentsReasoning
ollama run mixtral:8x7b
8GB: RAM offload12GB: RAM offload24GB: RAM offload

MoE local reasoning/chat when enough VRAM is available

Open model page -> Calculator keeps hardware-specific checks.

Llama 3.1 70B Instruct

70B Q4 about 44 GB Meta CodingAgentsReasoning
ollama run llama3.1:70b
8GB: Too large12GB: Too large24GB: RAM offload

High-quality local chat and reasoning on workstation-class hardware

Open model page -> Calculator keeps hardware-specific checks.

Qwen3-Next 80B-A3B Instruct

80B Q4 about 50 GB Alibaba CodingAgentsReasoning
ollama run qwen3-next:80b-a3b-instruct-q4_K_M
8GB: Too large12GB: Too large24GB: RAM offload

High-end local reasoning, long-context planning, and tool-agent workloads when 48GB+ memory is available

Open model page -> Calculator keeps hardware-specific checks.

Qwen3-Coder-Next

80B Q4 about 52 GB Alibaba CodingAgentsChat
ollama run qwen3-coder-next:q4_K_M
8GB: Too large12GB: Too large24GB: RAM offload

High-end local coding agents, repository-scale code edits, and tool-calling development workflows

Open model page -> Calculator keeps hardware-specific checks.

gpt-oss 120B

120B Q4 about 65 GB OpenAI CodingAgentsReasoning
ollama run gpt-oss:120b
8GB: Too large12GB: Too large24GB: RAM offload

Large local reasoning servers, heavy agent orchestration, and high-end homelab inference

Open model page -> Calculator keeps hardware-specific checks.

Devstral 2 123B

123B Q4 about 75 GB Mistral AI CodingAgentsReasoning
ollama run devstral-2:123b
8GB: Too large12GB: Too large24GB: RAM offload

High-end local software engineering agents, large-repository navigation, and tool-heavy coding workflows on server-class memory

Open model page -> Calculator keeps hardware-specific checks.

GLM-5.2

744B Q4 about 466 GB Z.ai CodingAgentsReasoning
ollama run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M
8GB: Too large12GB: Too large24GB: Too large

Long-horizon coding, large-repository context, frontend generation, and local sovereignty experiments on very large-memory systems

Open model page -> Calculator keeps hardware-specific checks.

Full Ollama command table

ModelCommandQ4 estimateCommon GPU fitCalculator
Qwen3 4B Thinking 2507ollama run qwen3:4b-thinking-2507-q4_K_M3.2 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Gemma 3 4Bollama run gemma3:4b3.5 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Qwen2.5 Coder 7Bollama run qwen2.5-coder:7b5.5 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Mistral 7Bollama run mistral:7b5.5 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Llama 3.1 8B Instructollama run llama3.1:8b6 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Qwen3 8Bollama run qwen3:8b6 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
DeepSeek-R1-0528-Qwen3-8Bollama run deepseek-r1:8b6 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Qwen3-VL 8B Instructollama run qwen3-vl:8b6.5 GB Q4 estimate8GB: Runs locally 12GB: Runs locally 24GB: Runs locallyOpen calculator
Qwen3.5 9Bollama run qwen3.5:9b6.6 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
Gemma 3 12Bollama run gemma3:12b9 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
Gemma 4 E4Bollama run gemma4:e4b9.6 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
Qwen2.5 Coder 14Bollama run qwen2.5-coder:14b10.5 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
DeepSeek R1 Distill Qwen 14Bollama run deepseek-r1:14b10.5 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
Phi-4 14Bollama run phi4:14b10.5 GB Q4 estimate8GB: RAM offload 12GB: Runs locally 24GB: Runs locallyOpen calculator
gpt-oss 20Bollama run gpt-oss:20b14 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Devstral Small 2 24Bollama run devstral-small-215 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3.5 27Bollama run qwen3.5:27b17 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3.6 27Bollama run qwen3.6:27b17 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Gemma 3 27Bollama run gemma3:27b18 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3 30B-A3B Instruct 2507ollama run qwen3:30b-a3b19 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3-Coder 30B-A3Bollama run qwen3-coder:30b19 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
GLM-4.7-Flashollama run glm-4.7-flash:q4_K_M19 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Gemma 4 31Bollama run gemma4:31b20 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3-VL 30B-A3B Instructollama run qwen3-vl:30b20 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen2.5 Coder 32Bollama run qwen2.5-coder:32b21 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
DeepSeek R1 Distill Qwen 32Bollama run deepseek-r1:32b21 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: Runs locallyOpen calculator
Qwen3.6 35B-A3Bollama run qwen3.6:35b-a3b24 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: RAM offloadOpen calculator
Mixtral 8x7Bollama run mixtral:8x7b28 GB Q4 estimate8GB: RAM offload 12GB: RAM offload 24GB: RAM offloadOpen calculator
Llama 3.1 70B Instructollama run llama3.1:70b44 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: RAM offloadOpen calculator
Qwen3-Next 80B-A3B Instructollama run qwen3-next:80b-a3b-instruct-q4_K_M50 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: RAM offloadOpen calculator
Qwen3-Coder-Nextollama run qwen3-coder-next:q4_K_M52 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: RAM offloadOpen calculator
gpt-oss 120Bollama run gpt-oss:120b65 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: RAM offloadOpen calculator
Devstral 2 123Bollama run devstral-2:123b75 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: RAM offloadOpen calculator
GLM-5.2ollama run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M466 GB Q4 estimate8GB: Too large 12GB: Too large 24GB: Too largeOpen calculator