self-hosted/ai
§01·model · /models

DeepSeek V3

llmactiveDeepSeek License Agreement

671B Mixture-of-Experts LLM by DeepSeek, roughly 37B active per token — its flagship open MoE for chat, code and reasoning. It is catalogued here as a reference point rather than a local option: nothing in this catalogue's 8-to-64 GB range can host it, and it carries no recipes. DeepSeek License Agreement.

§02·GPUs that run this model
0 total
No community benchmarks or recipes for this model yet.