Guides

Source-aware local AI planning guides

Use these guides to compare local hardware planning, cloud testing, and workflow tradeoffs without turning draft assumptions into buying advice.

Published planning guides

Published guideCloud vs Local planning

Cloud GPU vs Local GPU for AI WorkloadsSource-aware planning guide for deciding when to test on cloud GPU first and when local GPU planning may make more sense.Read guide →

Published guideLocal vs SaaS planning

Local AI vs AI SaaS for Workflow PlanningSource-aware planning guide for deciding when to run AI locally, test cloud GPU, or use AI SaaS/API tools.Read guide →

Published guideImage generation planning

Image Generation VRAM Planning for SDXL, SD3.5, and FLUXSource-aware guide for planning GPU VRAM around SDXL, Stable Diffusion 3.5 Large, and FLUX image-generation workflows.Read guide →

Published guideVRAM tier planning

12GB vs 16GB VRAM for Local AISource-aware guide for choosing a 12GB or 16GB local AI planning tier across dense LLMs, image generation, and MoE estimates.Read guide →

Published guideLocal LLM GPU planning

How to Choose a GPU for Local LLMsSource-aware guide for choosing a local LLM GPU planning path from model size, quantization, context, MoE behavior, and VRAM tiers.Read guide →

How to use the guide hub

Size

Estimate workload size

Start with VRAM planning before assuming a local or cloud route will fit.

Context

Review hardware context

Check GPU profiles, build routes, and comparisons before narrowing an option set.

Tradeoffs

Use guides for tradeoffs

Guide pages help compare control, setup effort, and workflow direction without provider ranking.

Verify

Verify the exact path

Validate your runtime, storage, privacy, and operational requirements before committing.

Model VRAM requirement pages

Start here when the question is model-specific: how a dense 7B or 8B model maps to planning tiers before runtime validation.

Model VRAMLlama

Meta Llama 3.1 8B InstructSource-backed LLM profile for long-context local planning.Open model page →

Model VRAMQwen

Qwen2.5 7B InstructSource-backed Qwen profile for calculator-ready local LLM planning.Open model page →

Model VRAMMistral

Mistral 7B Instruct v0.3Source-backed Mistral profile for calculator planning.Open model page →

Cloud GPU planning route

Planned guide topics

These topics remain in the planning backlog until the route, data quality, and source-backed scope are ready.

01How much VRAM do you need for AI workloads

Planning backlog item.

02AI workstation software stack planning