DeepSeek V4 Pro (DeepSeek-V4-Pro-0813) has gone live on the company's API, bringing enhanced agent c...

DeepSeek V4 Pro (DeepSeek-V4-Pro-0813) has gone live on the company's API, bringing enhanced agent capabilities that the company says approach or exceed those of Anthropic's Fable 5. The model supports a 1M-token context window and 384K-token maximum output, with both thinking and non-thinking modes. Pricing is set at RMB 3 per million input tokens, RMB 6 per million output tokens, and RMB 0.025 per million cached input tokens (approximately $0.42, $0.84, and $0.0035 respectively). The release follows the earlier V4 Flash update, which had signaled a Pro version would follow.