DeepSeek opens V4 Preview with 1M context and API compatibility

DeepSeek has opened V4 Preview with two new models, V4-Pro and V4-Flash, both offering a 1 million token context window, open weights, and same-day API access. The real enterprise signal is not just the scale. It is that DeepSeek is also lowering migration friction by supporting both OpenAI ChatCompletions and Anthropic APIs.

What’s new

In the official release, DeepSeek positions V4-Pro as a 1.6 trillion total parameter model with 49 billion active parameters, while V4-Flash is the lighter and cheaper variant at 284 billion total and 13 billion active parameters. Both support thinking and non-thinking modes, and both are available immediately in DeepSeek’s own services and API.

DeepSeek is also publishing open weights on Hugging Face and has announced that the older deepseek-chat and deepseek-reasoner models will be fully retired on July 24, 2026. That makes this more than a demo. It is the start of a real migration path.

Why this matters

This matters for enterprises because DeepSeek is trying to reduce two of the biggest adoption barriers at the same time: context length and integration friction. A 1M window is useful for document-heavy, code-heavy, and tool-driven workflows, but the more important detail may be API compatibility. If companies can test or swap in DeepSeek without rewriting the whole application layer, pricing pressure on closed model vendors goes up.

For CIOs, that means more optionality in the stack, especially for internal agents, developer tooling, and workflows where open weights and lower cost matter more than brand assurance.

Source and date validation

The original announcement is on DeepSeek’s official API documentation in the page news260424, indicating April 24, 2026, and stating that the API is updated and available today. Hugging Face also published a DeepSeek V4 analysis post on April 24, 2026, and DeepSeek’s official Hugging Face collection was updated the same day. This is within the 48-hour window.

Sources:

• https://api-docs.deepseek.com/news/news260424
• https://huggingface.co/collections/deepseek-ai/deepseek-v4
• https://huggingface.co/blog/deepseekv4