Phi-4
Best Vision Models You Can Run Locally: Every Model, Every GPU Tier (2026)
Qwen3-VL 8B replaced Qwen2.5-VL as the best local vision model. Full VRAM table, Ollama commands, speed benchmarks, and setup for every GPU from 4GB to 48GB+. Updated March 2026.
Best Local LLMs for Math & Reasoning: What Actually Works
The best local LLMs for math and reasoning tasks, ranked by VRAM tier. AIME and MATH benchmarks for DeepSeek R1, Qwen 3 thinking, and Phi-4-reasoning.