DeepSeek-V3 (Mar 2025)
DeepSeekLanguage modeling/generationCode generationQuantitative reasoningQuestion answeringOpen weights
Developed by DeepSeek in 2025, DeepSeek-V3 (Mar 2025) is a language modeling/generation model with 671000000000.0 parameters with openly available weights.
About DeepSeek-V3 (Mar 2025)
We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) an
LLM pricing & performance
Full LLM page →DeepSeek-V3 (Mar 2025) is available via API — live cost, context, and benchmark data:
Input / 1M
—
Output / 1M
—
Context
—
Tokens/sec
—
Aider Polyglot: 55.1LMArena (Chatbot Arena) Elo: 1332.5
Details
- Provider
- DeepSeek
- Task
- Language modeling/generation,Code generation,Quantitative reasoning,Question answering
- Parameters
- 671000000000.0
- Released
- 2025-03-24
- Open weights
- Yes