DeepSeek V4 Pro is live on Clusterbase
An open model with a million-token context and reasoning on by default, hosted by Clusterbase in the US with zero data retention.
DeepSeek V4 Pro is now available in Cluster and to developers through Clusterbase. It is an open model with a million-token context, hosted by us in the United States. Nothing you send is kept.
This is the release build of V4 Pro, not the preview. DeepSeek published the 0813 build this month under the MIT license, and it is the first DeepSeek model on Clusterbase. Like every open model here, it has to be good at real work, and the privacy boundary has to be simple. Requests are processed with zero data retention. Your prompts and outputs are not stored for training or kept after the request is served.
A large context window and a large answer
DeepSeek V4 Pro accepts up to 1,048,576 tokens of context and can produce up to 393,216 tokens in a response. That is room for a whole codebase or a long research thread on the way in. On the way out, it is room for a full refactor, a long report, or a complete set of files, without splitting the job into pieces first.
The model is text only. It does not accept images or PDFs. If your work needs those, Kimi K3 and Qwen 3.8 Max are in the same picker.
Reasoning is on by default
DeepSeek V4 Pro reasons before it answers. You can set the effort to none,
high, or max; Cluster uses high by default. Turn it off when you want a
fast, direct answer. Leave it on for difficult coding, analysis, and long
multi-step tasks.
Built for agentic work
The release build beats the preview on every published agentic benchmark, the tasks where the model has to operate tools, navigate a repository, and finish a job on its own. On DeepSWE the score went from 12.8 to 62.7.
Scores published in the DeepSeek V4 Pro 0813 model card. These are not Clusterbase measurements.
Use it where you already work
In Cluster, choose DeepSeek V4 Pro from the model picker on the web or mobile. In Cluster Build, select it from the model list in your terminal.
Developers can use the same model through Clusterbase's OpenAI-compatible API
with the model ID deepseek-v4-pro-0813. Existing OpenAI-compatible clients
only need the Clusterbase endpoint, credentials, and model name. See
Models and pricing for the rate card and the
LLM Gateway API reference for the request shape.
DeepSeek V4 Pro does not replace the Smart or Fast model choices. It is an explicit option for the work where you want a large context window, long answers, deep reasoning, and a simple privacy boundary: it runs in the US, and nothing you send is kept.