DeepSeek Drops V4 With 1 Million Token Memory, And Its Competing With ChatGPT

DeepSeek has launched V4, its boldest AI model yet, and it's gunning for ChatGPT without access to America's most powerful chips. Here's what makes it different.

DeepSeek's new V4 model nearly matches Google's Gemini on key tests, all while working around US chip restrictions.

Source : Unsplash

Show Quick Read

Key points generated by AI, verified by newsroom

• DeepSeek unveils V4 model, boosting AI performance and reasoning.
• V4 offers ultra-long context window, processing vast amounts of data.
• Pro version nears top closed systems, Flash prioritizes speed.

China’s artificial intelligence race is heating up, and a new contender has just stepped into the spotlight. DeepSeek has unveiled its latest model, DeepSeek-V4, claiming major gains in performance, efficiency, and reasoning abilities. The company says a preview version is already available, giving users an early look at what it describes as one of the most capable open-source AI systems so far. 

The release also reflects China’s growing push to build advanced AI tools despite ongoing restrictions on access to high-end chips and global competition.

What Makes DeepSeek-V4 Stand Out In The AI Race?

DeepSeek says its V4 model comes with an “ultra-long context” of up to one million tokens, a sharp jump from the 128,000-token limit seen in its earlier version. This means the model can process and understand significantly larger amounts of information at once, including entire books or large codebases.

The company has launched two variants: DeepSeek-V4-Pro and DeepSeek-V4-Flash. The Pro version is positioned as the more powerful option, while Flash is designed to be faster and more cost-efficient. 

According to DeepSeek, V4-Pro performs strongly across benchmarks, even coming close to top closed systems like Gemini-Pro-3.1.

It also introduces a “maximum reasoning effort mode”, which the company claims boosts its ability to handle complex tasks. However, DeepSeek acknowledges that it still trails models like Claude Opus 4.6 in some areas.

How Chip Restrictions And Efficiency Shape DeepSeek’s Strategy?

The launch comes at a time when China faces strict US export controls on advanced GPUs, crucial for training AI systems. DeepSeek has not revealed the exact chips used for V4, but says the model is compatible with both Nvidia and Huawei hardware.

This focus on optimisation is key. The company claims a “dramatic leap” in computational efficiency, allowing faster processing and lower costs. Its earlier R1 model had already shaken markets by offering performance comparable to systems like ChatGPT at a fraction of the cost.

With V4, DeepSeek is aiming to push further, improving real-world usability while strengthening China’s position in the global AI landscape.

Claude Mythos AI Sparks Panic: FM Sitharaman Flags Bank Security Risks

Meta Is Recording Employee Mouse Moves To Build AI That Does Your Job