DeepSeek Launches V4 Pro AI Model with 1.6 Trillion Parameters and New Pricing Structure
DeepSeek has officially launched its V4 Pro model, introducing one of the largest open-weight AI systems to the market. This model boasts a staggering 1.6 trillion parameters, with 49 billion of them active per token, positioning it as a formidable competitor to proprietary models from OpenAI, Google, and Anthropic.
What sets V4 Pro apart is its availability under an MIT license, allowing users to modify and deploy the model freely. This open-weight approach enables developers to fine-tune the model for specific applications, such as healthcare or fintech, without incurring per-token fees.
The V4 Pro model utilizes a Mixture-of-Experts (MoE) architecture, which means only a fraction of its total parameters are active at any given time, optimizing performance for various tasks. It supports a context window of up to 1 million tokens and has been integrated across DeepSeek's web platform, mobile app, and API.
In addition to the launch, DeepSeek announced a significant price increase for its API services, effective August 16. The output token price for the V4 Pro model will rise from $0.87 to $3.96 per million tokens during peak hours, marking a 4.5x increase. This shift indicates a strategic move towards aligning with industry pricing standards after previously undercutting competitors.
DeepSeek's pricing adjustments come after a period of aggressive competition, where the company had slashed its prices, prompting other AI firms in China to follow suit. The new pricing structure introduces peak and off-peak billing, which may lead developers to reconsider their token usage and adapt their applications accordingly.
FAQ
What is the V4 Pro AI model launched by DeepSeek?
The V4 Pro AI model is one of the largest open-weight AI systems, featuring 1.6 trillion parameters, with 49 billion active per token. It is designed to compete with proprietary models from companies like OpenAI, Google, and Anthropic.
What are the key features of the V4 Pro model?
Key features of the V4 Pro model include its Mixture-of-Experts (MoE) architecture, which activates only a fraction of its parameters at a time, a context window of up to 1 million tokens, and its availability under an MIT license for free modification and deployment.
How has DeepSeek changed its pricing structure for API services?
Effective August 16, the output token price for the V4 Pro model will increase from $0.87 to $3.96 per million tokens during peak hours, marking a 4.5x increase. The new pricing structure also introduces peak and off-peak billing.
Why did DeepSeek decide to increase its API pricing?
DeepSeek's pricing increase is a strategic move to align with industry pricing standards after a period of aggressive competition, where they had previously undercut competitors' prices.
How can developers utilize the V4 Pro model for their applications?
Developers can utilize the V4 Pro model by modifying and deploying it freely under the MIT license, allowing them to fine-tune the model for specific applications such as healthcare or fintech without incurring per-token fees.
Comments
Comments are moderated before publish.
No comments yet — be the first.