DeepSeek · LLM
DeepSeek V3
671B MoE open-weight base model.
Capabilities
- ▸MoE
- ▸open weights
Use cases
self-hosting · coding
Lineage
View the full DeepSeek lineage →DeepSeek V3 head-to-head comparisons
- DeepSeek-V3 vs Llama 3.3 70BBoth are open-weight 128K models. DeepSeek-V3 is a much larger mixture-of-experts model; Llama 3.3 70B is a dense model that is far easier to self-host.
- Llama 4 vs DeepSeek V3The two most-deployed open-weight families of 2025: Llama 4 offers an enormous context window and a broad ecosystem, DeepSeek V3 delivers strong reasoning per dollar.
Go deeper
Sources & verification
Information is compiled from official announcements, documentation and reputable industry coverage.
Last verified Sep 10, 2026