Chinese AI company DeepSeek caused quite a stir last year as it topped the App Store charts and overtook incumbents like ChatGPT. That was a year ago and now the company has launched a preview of its latest and more capable model, DeepSeek-V4.
If you have a DeepSeek account, you can try the new model here: chat.deepseek.com/. There are two editions available, Expert and Instant. Expert is a huge 1.6 trillion parameter model, Instant is a more reasonable (but still huge) 284 billion parameter model.
DeepSeek-V4 Pro and Flash models
And here’s the most interesting part – it’s an open-source model, meaning you can download it from Hugging Face and run it on your own hardware at home. Well, you could say that theoretically – you’ll need some serious hardware to actually run these models.
Now, the Expert (Pro) model has 49 billion active parameters and the Instant (Flash) has only 13 billion active parameters – these are the parameters that actually need to fit in VRAM. However, shuffling parameters between VRAM and system memory slows down token generation. The nice thing about open-weight models is that the community can create quantified and distilled models that can actually run on consumer-level hardware.
Anyway, the company says that DeepSeek-V4-Flash, the model that powers Instant Mode, has reasoning capabilities close to V4-Pro and performs equally well on simple agentic tasks. Both models have a 1 million token reference window.
deepseek-v4-flash benchmark scores
According to DeepSeek testing, DeepSeek-v4-Pro, which is behind Expert Mode, has the best reasoning skills for math, STEM, and coding tasks among open-source models and rivals the performance of closed-source models. The company says V4-Pro outperforms most AI models in terms of world knowledge, losing only to Google’s Gemini-3.1-Pro.
DeepSeek-V4-Pro vs. Its Major Competitors
DeepSeek-V4 can be used with popular tools such as CloudCode, OpenCloud, and OpenCode. If you can’t run the model at home, you can use the company’s API. Here’s the pricing:
DeepSeek-V4 Pro and Flash API Pricing