DeepSeek V4 Still Awaiting Public Launch Despite Persistent April Release Predictions

As of April 6, 2026, DeepSeek's highly anticipated V4 model has not yet launched publicly, despite mounting industry speculation and credible reports pointing to an imminent release. According to a Reuters report from April 3, citing The Information, the model is likely to launch in the 'next few weeks' and will run on Huawei's newest chips. This follows months of work where DeepSeek engineers collaborated with Huawei and Cambricon to rewrite parts of the model stack and conduct extensive testing. Crucially, the company's public API documentation still only lists the `deepseek-chat` and `deepseek-reasoner` endpoints, both explicitly mapped to the DeepSeek-V3.2 architecture with a 128K context limit, confirming V4 is not yet live for general API use.

This launch cycle has been marked by shifting timelines, moving from initial rumors of a February or March release to the current 'near-term' expectation. Key reports throughout early 2026 have painted a picture of a model focused on code generation and extremely long coding prompts, with DeepSeek notably not providing early access to Nvidia or AMD for optimization, instead prioritizing domestic suppliers like Huawei. For developers and enterprises, the distinction between 'launching soon' and 'publicly available' is critical; integration decisions and production planning must still treat V4 as an unreleased asset until official model IDs and pricing appear in the API docs.