DeepSeek releases V4 model with one million token context

Read the original article →

What happened

Chinese AI lab DeepSeek released its V4 model. The headline feature is an ultra-long context window of one million tokens. That lets the model take in very large inputs at once, such as long documents, big codebases, or many files.

DeepSeek says V4 reaches performance close to leading closed source models while costing less to run. The release is part of a busy year for Chinese labs, which have been shipping new models at a fast pace.

Why it matters

Long context is useful for real work. With a one million token window, a model can read and reason over a whole project or a stack of reports without losing track. Doing this at lower cost matters for teams that process large amounts of text.

DeepSeek's progress also shows how quickly Chinese labs have closed the gap with US models on price and capability.

MintedBrain take

For users, the main draw is cost and context. If you work with long documents or large codebases, a cheap model with a huge context window can save real money. As always, check data and privacy rules before sending sensitive content to any model.

References

This article was originally published at CNBC. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News