Google ships four Gemini Flash models in 106 days, but flagship Pro remains delayed

Google released Gemini 3.8 Flash on Wednesday, its fourth Flash model in 106 days, but its flagship Gemini 3.5 Pro remains delayed. The new model matches Anthropic's Claude Opus 5 on coding benchmarks at roughly one-fifth the cost. Gemini 3.8 Flash excels at coding and agentic tasks, passing 74% of DeepSWE runs at an average cost of $2.36 per task versus Opus's $11.84. Google attributes the rapid Flash cadence to recursive self-improvement loops, where models refine their own code, though the Pro model's June release has slipped. Google's best model now ranks 10th on the Artificial Analysis Intelligence Index, trailing rivals like Meta's Muse Spark 1.3. The company has invested heavily in coding, with cofounder Sergey Brin urging teams to make models "primary developers," while commercial demand for cheaper Flash APIs grows amid constrained computing supply.