DeepSeek released its flagship V4-Pro model (build 0813) on August 12, 2026, ending a preview period that began earlier in the year. The mixture-of-experts system uses 1.6 trillion total parameters with 49 billion active per token, supports a context window of up to one million tokens, and emphasizes agentic capabilities such as tool use and multi-step task execution. DeepSeek reported strong scores on agent-focused benchmarks including Terminal Bench 2.1 and DeepSWE, and confirmed a price increase for output tokens is set to follow the general-availability launch.
