TSMC's July sales rose 45% year-over-year, driven by sustained AI chip demand that continues to outpace tariff-related headwinds, according to the report. The article, however, focuses primarily on DeepSeek's new V4-Flash model, described as the cheapest frontier-grade AI ever benchmarked by Artificial Analysis, costing about three cents to run a full test suite. DeepSeek's official API docs list V4-Flash at $0.14 per million cache-miss input tokens, $0.28 per million output tokens, and $0.0028 per million cache-hit input tokens. The company's pricing page says product prices may vary and reserves the right to adjust them, but it does not give a date for a new price or publish a fresh August warning about an imminent hike. The article also refutes claims that a founder named Jun Song defended a 2x to 10x increase on X, noting DeepSeek's widely reported founder is Liang Wenfeng. According to DeepSeek's April 24 change log, the API began supporting DeepSeek-V4-Pro and V4-Flash through OpenAI-compatible and Anthropic-compatible interfaces, with old model names mapping to V4-Flash compatibility modes until July 24, 2026. Artificial Analysis lists V4-Flash as an April 2026 release with 284 billion total parameters and 13 billion active parameters, scoring 40 on its Intelligence Index and generating output