DeepSeek Unveils 3V Model, The Most Powerful Open-Source AI Yet

The Chinese AI company DeepSeek released its latest open-source model, DeepSeek-V3, this week. The startup claims that its new large model includes 671B parameters and can perform better than frontier models like OpenAI’s GPT-4o and Anthropic’s Claude-Sonnet-3.5.

Users can interact with the model through the official website, read the company’s paper, and access the model via Hugging Face. DeekSeek said this new model is three times faster than the previous version released in November and has enhanced capacities and API compatibility.

🚀 Introducing DeepSeek-V3!

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

🐋 1/n pic.twitter.com/p1dV9gJ2Sd

— DeepSeek (@deepseek_ai) December 26, 2024

“We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token,” wrote the company on GitHub. “Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models.”

Many users, organizations, and companies have congratulated DeepSeek on its achievement. “Congratulations on the stellar release!” wrote Hugging Face on X, “The model checkpoints and a detailed report – truly Christmas is here!”

Andrej Karpathy—former researcher at OpenAI and head of AI at Tesla—shared a few thoughts too. “DeepSeek (Chinese AI co) making it look easy today with an open weights release of a frontier-grade LLM trained on a joke of a budget (2048 GPUs for 2 months, $6M).”

DeepSeek-V3 is now one of the strongest open-source AI models in the market and its paid API remains among the most affordable in the market.