DeepSeek-V2
by DeepSeek · open-source · Last verified 2026-03-17
DeepSeek's mixture-of-experts model introducing Multi-head Latent Attention (MLA) for dramatically reduced inference cost. Activates 21B of its 236B total parameters per token while matching larger dense models.
https://www.deepseek.com ↗B
B—Above Average
Adoption: BQuality: B+Freshness: C+Citations: B+Engagement: F
Specifications
- License
- DeepSeek License
- Pricing
- open-source
- Capabilities
- text-generation, code-generation, reasoning, multilingual, efficient-inference
- Integrations
- huggingface, vllm, ollama
- Use Cases
- cost-efficient-inference, chatbots, code-generation, research, enterprise-deployment
- API Available
- Yes
- Parameters
- 236B (21B active)
- Context Window
- 128K tokens
- Modalities
- text
- Training Cutoff
- Early 2024
- Tags
- llm, open-source, moe, cost-efficient, multi-head-latent-attention, deepseek
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
60.8Adoption
68
Quality
78
Freshness
55
Citations
72
Engagement
0