DeepSeek V4 Pro Released
DeepSeek V4 Pro launches with 1M context, 384K output, low pricing, and Agent focused API support.
“AI Disruption” Publication 10,400 Subscriptions 20% Discount Offer Link.
One day, three heavyweight models rarely collided. Grok 4.6’s paper scores approached Fable 5.
The total-parameter 2.4T Qwen3.8 Max also opened its weights as scheduled. On the DeepSeek side, the official version of DeepSeek V4 Pro (deepseek-v4-pro) launched simultaneously and opened its API.
The latest model and pricing page show that the model version corresponding to deepseek-v4-pro has been updated to DeepSeek-V4-Pro-0813.
From the information currently released, deepseek-v4-pro supports a 1M context length, with a maximum output length of 384K tokens. It supports both non-thinking mode and thinking mode, with thinking mode enabled by default.
For long codebases, large-scale document processing, and Agent tasks that require continuous multi-step execution, the 1M context and 384K maximum output mean that a single call can accommodate and generate a substantial amount of content.
Its supporting capabilities also basically cover the mainstream requirements for Agent development at present.
deepseek-v4-pro supports JSON Output, Tool Calls, the Responses API, and the Anthropic API. It also supports Beta-stage conversation prefix continuation and FIM completion, with FIM completion limited to non-thinking mode only.
API access is compatible with both OpenAI and Anthropic formats.
