Model requirements
Use this index to choose a local model by practical VRAM tier, use case, Ollama command, and common GPU fit. Open a model page when you need the full GPU table.
Start with 4B-8B Q4 models. Look for green 8GB badges.
Good floor for local agents and 7B/14B models with headroom checks.
Use for larger coding models, longer context, and heavier local workflows.
Small local reasoning, routing, tool decisions, and lightweight coding on 6GB-8GB GPUs
qwen3:4b-thinking-2507-q4_K_M
Open calculator ->
Small multimodal local assistant and low-resource setups
gemma3:4b
Open calculator ->
Small local coding assistant and agent tool generation
qwen2.5-coder:7b
Open calculator ->
Fast local chat and simple agent tasks
mistral:7b
Open calculator ->
Fast local chat, lightweight agents, low-cost local testing
llama3.1:8b
Open calculator ->
Fast general local assistant with reasoning/coding balance
qwen3:8b
Open calculator ->
Updated local reasoning experiments, coding logic checks, and step-by-step technical analysis
deepseek-r1:8b
Open calculator ->
Local image understanding, OCR-style document triage, screenshots, and lightweight multimodal agent routing
qwen3-vl:8b
Open calculator ->
Modern multimodal local assistant, agent experiments, and coding support on mainstream GPUs
qwen3.5:9b
Open calculator ->
Balanced multimodal local chat on 12GB+ GPUs
gemma3:12b
Open calculator ->
Efficient multimodal local assistant and edge-style agent workflows
gemma4:e4b
Open calculator ->
Local coding, scripts, repo assistance, technical agents
qwen2.5-coder:14b
Open calculator ->
Local reasoning and debugging on 12GB/16GB GPUs
deepseek-r1:14b
Open calculator ->
Compact reasoning and technical assistant on 12GB/16GB GPUs
phi4:14b
Open calculator ->
Local reasoning, agent planning, and tool-use workflows on 16GB+ GPUs
gpt-oss:20b
Open calculator ->
Software engineering agents, repo navigation, patch planning, and local coding workflows
devstral-small-2
Open calculator ->
24GB-class multimodal agent, coding assistant, and reasoning workloads
qwen3.5:27b
Open calculator ->
Newest 27B-class local multimodal coding, agent, and reasoning workloads on 24GB-class GPUs
qwen3.6:27b
Open calculator ->
High-quality multimodal local assistant on 24GB GPUs
gemma3:27b
Open calculator ->
General local reasoning, tool agents, multilingual writing, and coding on 24GB+ GPUs
qwen3:30b-a3b
Open calculator ->
Agentic coding, repository-scale local code review, and tool-heavy development loops
qwen3-coder:30b
Open calculator ->
Local coding agents, terminal workflows, tool-heavy engineering tasks, and 30B-class reasoning on 24GB+ GPUs
glm-4.7-flash:q4_K_M
Open calculator ->
High-quality multimodal reasoning, coding assistants, and local-first agent workflows
gemma4:31b
Open calculator ->
Higher-quality local multimodal reasoning, screenshot analysis, document/image extraction, and GUI-agent planning
qwen3-vl:30b
Open calculator ->
Strong local coding and architecture work on 24GB GPUs
qwen2.5-coder:32b
Open calculator ->
Heavy local reasoning on 24GB GPUs
deepseek-r1:32b
Open calculator ->
Agentic coding and multimodal reasoning when 27B is not enough and 32GB-class headroom is available
qwen3.6:35b-a3b
Open calculator ->
MoE local reasoning/chat when enough VRAM is available
mixtral:8x7b
Open calculator ->
High-quality local chat and reasoning on workstation-class hardware
llama3.1:70b
Open calculator ->
High-end local reasoning, long-context planning, and tool-agent workloads when 48GB+ memory is available
qwen3-next:80b-a3b-instruct-q4_K_M
Open calculator ->
High-end local coding agents, repository-scale code edits, and tool-calling development workflows
qwen3-coder-next:q4_K_M
Open calculator ->
Large local reasoning servers, heavy agent orchestration, and high-end homelab inference
gpt-oss:120b
Open calculator ->
High-end local software engineering agents, large-repository navigation, and tool-heavy coding workflows on server-class memory
devstral-2:123b
Open calculator ->
Long-horizon coding, large-repository context, frontend generation, and local sovereignty experiments on very large-memory systems
hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M
Open calculator ->