DeepSeek's New AI Is Faster and Cheaper – Testers Wanted

DeepSeek just opened testing for an updated version of its Flash AI model, called V4.1 Flash. This is an in-between release, not the final version, but developers are already being invited to try it through the same digital doorway (API) they use today. The big selling point is that it is built on a new design that understands not just text but also images – what tech folks call native multimodal support. In plain English, it means the AI can look at a picture, read a chart, and answer questions about them without extra steps.

Why should you care? Because two of the biggest complaints about AI are that it can be slow and expensive. DeepSeek says this new model is both faster and cheaper to run. That likely means apps, chatbots, and other services that use AI behind the scenes could become more responsive and cost less to maintain. Companies often pass those savings on to you, the user – or at least keep prices steady while improving performance.

Getting to try it is surprisingly simple for a developer: they keep the same connection address (called a base_url) and just switch the model name to a special code that includes an expiration date: deepseek-v4.1-flash-expires-on-0910. Pricing stays identical to the current Flash model, and each account can make up to 20 requests at the same time. That tells you this is aimed at real-world testing rather than huge launches.

One honest catch: this is a test version, so there will likely be bugs and rough edges. The "expires" part in the model name is a reminder that this version won't stick around forever. Still, the move shows how quickly AI models are improving. Faster, cheaper, and able to handle images could mean your favorite AI tools get better and more affordable in the near future.