# 🚀 Introducing DeepSeek-V3

## Biggest leap forward yet

- ⚡ 60 tokens/second (3x faster than V2!)
- 💪 Enhanced capabilities
- 🛠 API compatibility intact
- 🌍 Fully open-source models & papers

![](https://cdn.deepseek.com/api-docs/ds_v3_tps_en.gif)

![](https://cdn.deepseek.com/api-docs/ds_v3_benchmark_table_en.jpeg)

***

## 🎉 What’s new in V3

- 🧠 671B MoE parameters
- 🚀 37B activated parameters
- 📚 Trained on 14.8T high-quality tokens

🔗 Dive deeper here:

- Model 👉 <https://github.com/deepseek-ai/DeepSeek-V3>
- Paper 👉 <https://github.com/deepseek-ai/DeepSeek-V3/blob/main/DeepSeek_V3.pdf>

***

## 💰 API Pricing Update

- 🎉 Until Feb 8: same as V2!
- 🤯 From Feb 8 onwards:

| Input (cache miss) | Input (cache hit) | Output         |
| ------------------ | ----------------- | -------------- |
| $0.27/M tokens     | $0.07/M tokens    | $1.10/M tokens |

## Still the best value in the market! 🔥

![](https://cdn.deepseek.com/api-docs/ds_v3_price_en.jpeg)

![](https://cdn.deepseek.com/api-docs/ds_v3_price_2_en.jpeg)

***

- 🌌 Open-source spirit + Longtermism to inclusive AGI
- 🌟 DeepSeek’s mission is unwavering. We’re thrilled to share our progress with the community and see the gap between open and closed models narrowing.
- 🚀 This is just the beginning! Look forward to multimodal support and other cutting-edge features in the DeepSeek ecosystem.
- 💡 Together, let’s push the boundaries of innovation!
