Written by Jakub Rusinowski · Last updated July 10, 2026
DeepSeek's groundbreaking non-reasoning frontier model. Trained for just $5.5M — 1/10th the cost of comparable proprietary models. Uses a 685B MoE architecture (37B active parameters per token), topping the open-source leaderboard for coding, math, and general instruction following at launch. Previous generation — superseded by DeepSeek V4, which expands context to 1M tokens and improves MoE efficiency. Still widely used and supported.
| DeepSeek V3 (685B MoE) | Min 400 GB VRAM · Q4_K_M · 128,000 ctx · ollama run deepseek-v3 |
The cheapest GPU that runs DeepSeek V3 locally (min 400 GB VRAM) is the Apple M3 Ultra (512 GB).
Install Ollama then run: ollama run deepseek-v3
Minimum VRAM: 400 GB. For best results use Q4_K_M quantization.
DeepSeek V3 needs about 400 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: DeepSeek V3 (685B MoE) (400 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
DeepSeek V3's smallest variant needs about 400 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
Q4_K_M is the best balance of quality and VRAM for DeepSeek V3 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run deepseek-v3. This downloads DeepSeek V3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.