GEMINI 3.8 FLASH SHOWS HOW FAST AI IS BEING COMMODITIZED $GOOGL launched Gemini 3.8 Flash for coding and agentic workflows alongside a Cyber model built specifically for vulnerability discovery and patching. Flash does not beat the frontier models everywhere, but it wins roughly half the benchmark set while costing just $0.75 per million input tokens and $3.75 per million output through year-end. That price-performance gap is really the point. Flash costs roughly one-seventh as much as Claude Opus 5 on output while remaining competitive across many workloads, which suggests the cheaper model tier is getting good enough for a much larger share of always-on agentic work rather than simply serving as a lightweight fallback. The catch is that the introductory pricing doubles in 2027 to $1.50 input and $7.50 output, so some of today’s cost advantage narrows quickly. That makes this less about Google permanently undercutting everyone and more about how fast capable models are moving down the cost curve. Cyber is where the story gets more interesting. Google says it achieved more than 70% success across 20 languages and produced 2.6x more correct patches than larger commercial models, but access is initially restricted to trusted defenders through Fairwind. That tells you Google sees the offensive capability as real too, and as these models get better at finding vulnerabilities the distribution question may become just as important as the benchmark itself.