Concepts
11 practical guides in Concepts, grounded in measured local hardware data.
- Maya model sizing: how much VRAM plus RAM a local LLM needs
GLM-5.3-Flash, DeepSeek-V4.1-Flash, MiniMax-M3.1-Flash and Qwen-3.8-Flash-Next need very different memory totals. The VRAM plus RAM table, the recommended sizes, and the floor below them.
- What Is Quantization in Local LLM Deployment for Strata
Learn how quantization reduces model size and RAM usage for local AI on Strata. Understand trade-offs in accuracy and speed with our guide.
- vram-vs-ram: Local AI Memory Guide
Learn why VRAM matters for local LLMs. Stratat uses RAM for large models. See how to check your hardware and optimize your Stratat client setup.
- Understanding tokens per second in Local LLM Inference
Learn how tokens-per-second measures local AI speed on your hardware. See how to calculate inference rates for Strata engine deployments without cloud costs.
- Mastering Your Context Window for Local LLMs
Understand context windows in local LLMs. Learn how to measure your VRAM, estimate costs at $0.17/kWh, and optimize your Strata client settings effectively.
- moe-explained: local LLM efficiency guide
Discover how Mixture of Experts optimizes local LLM inference on Stratat. Learn to measure token throughput and energy use on your desktop for better performance.
- Understanding KV Cache for Local LLMs
Learn how the KV-cache impacts latency in local LLMs on Strata. Explore memory usage and how to optimize your setup for faster generation speeds.
- What Are Open Weights in Local AI Deployment?
Understand open-weights in local AI. Strata runs these models locally. Learn how to verify model integrity and deploy efficiently on your desktop.
- Temperature Control in Local LLMs
Master temperature settings for your Stratat client to balance creativity and precision. Learn how to tweak this parameter for better local AI results.
- Understanding Memory Limits in Local LLMs
Learn why GPU memory bandwidth is the bottleneck for local LLMs like Strata. We explain how VRAM speed dictates inference rates and local deployment performance.
- Local vs Cloud LLMs: Cost and Privacy Tradeoffs
Discover how running AI locally with Strata saves money on API calls and protects your data. Compare latency, privacy, and hardware requirements to choose the right deployment.