DeepSeek V4 Flash (Reasoning, High Effort)

DeepSeek V4 Flash is a reasoning-focused AI model developed by DeepSeek, released on April 24, 2026. Designed for high-effort tasks, the model achieves an intelligence index of 37.5 and a coding index of 52. Its performance across various benchmarks includes a GPQA score of 0.867, an HLE score of 0.278, and a SciCode score of 0.42. Additional evaluation metrics include an IFBench score of 0.734693877551021, an LCR score of 0.626666666666667, a TerminalBench Hard score of 0.386363636363636, and a TAU2 score of 0.956140350877193. The model is available at a pricing structure of $0.14 per 1 million input tokens and $0.28 per 1 million output tokens, resulting in a blended cost of $0.17 per 1 million tokens.