DeepSeek V4 Pro 0813 (on OpenRouter)

DeepSeek's V4 Pro 0813 was first available only through an API, and Simon Willison's post initially said plans for open weights could not be confirmed. An update to that post points to the official model repository, which provides weights, an MIT license and instructions for running the model locally.

That resolves the original release-status uncertainty. Calling a hosted model and obtaining its weights are different forms of access; the repository makes operating a deployment yourself a documented option. The initial API-only description records an earlier stage of the release.

“The latest DeepSeek Pro model is now available, via API only. I had to link to OpenRouter because DeepSeek don't have any obvious announcement page for their new model.”

“I haven't been able to confirm if they plan to release the open weights, but given the weights are available for both April's deepseek-ai/DeepSeek-V4-Pro and July's deepseek-ai/DeepSeek-V4-Flash-0731 it seems likely.”

“Interestingly I got very different looking pelicans for the three different reasoning levels of low, medium, and high. I've not noticed this kind of difference from any other model”

“In terms of benchmarks... as far as I can tell those were released to the Official DeepSeek WeChat Group, then copied and pasted into a post on Reddit which was deleted by the moderators for being "low-effort", then copied into this ASCII-art table on Hacker News”