DeepSeek opens beta of V4.1 Flash with native multimodal support

DeepSeek has opened internal beta testing for an intermediate version of its V4.1 Flash model, according to a post on X relayed to r/LocalLLaMA. The beta adopts a new model architecture with native multimodal support, stronger capabilities, faster speeds, and lower costs, the post says.

Developers can call the model through the API by keeping their base_url unchanged and setting the model name to deepseek-v4.1-flash-expires-on-0910. Pricing is currently identical to deepseek-v4-flash, with a rate limit of 20 concurrent requests per account.

The announcement comes as DeepSeek continues to iterate quickly on its Flash line, which is positioned as a fast, low-cost tier. The "expires-on-0910" naming suggests the beta model name is temporary and tied to a specific cutoff date. The claims originate from a single X account and have not been confirmed by DeepSeek directly, so details such as the exact architecture changes and final pricing remain unverified.