FriendliAI Expands Model Offering With DeepSeek-V4 Integration on Dedicated Endpoints

According to a recent LinkedIn post from FriendliAI, the company is featuring DeepSeek AI’s new DeepSeek-V4-Pro and DeepSeek-V4-Flash models on its Friendli Dedicated Endpoints, with both reportedly offering a 1 million-token context window. The post highlights that these open-weight models have become among the most-used on OpenRouter within weeks of launch, suggesting growing developer interest in high-capacity, cost-efficient large language models.

The company’s LinkedIn post emphasizes that DeepSeek-V4-Flash is positioned as a performance-efficiency option, with 284 billion total parameters and 13 billion active per token, while achieving reasoning benchmark scores close to DeepSeek-V4-Pro and competitive with leading proprietary models. DeepSeek-V4-Pro is presented as a higher-capability tier, with strong scores in world knowledge, coding, and long-context benchmarks, targeting more demanding workloads.

For investors, the post suggests that FriendliAI is seeking to strengthen its role as an infrastructure provider for advanced open-weight models by integrating what it portrays as frontier-level systems into its dedicated endpoints. This alignment with popular, high-performing open-weight models could enhance platform stickiness, broaden the developer and enterprise user base, and potentially improve monetization through higher-throughput and premium workloads over time.

Strategically, offering both a high-throughput option (Flash) and a higher-ceiling model (Pro) may allow FriendliAI to address diverse use cases such as agents, coding assistants, and long-context enterprise applications. If adoption trends on OpenRouter translate into sustained usage on FriendliAI’s infrastructure, the company could benefit from volume-driven revenue growth and improved positioning in the competitive AI infrastructure and model-serving market.