[Google Gemini 3.8 Flash Launches, Scores 71% on DeepSWE 1.1] According to Deep潮 TechFlow, on September 2, Google Gemini 3.8 Flash began rolling out on Gemini, Google AI Studio, and API. Testing Catalog reports that the model scored 71% on the DeepSWE 1.1 benchmark, close to Claude Opus 5's 74%, but at a significantly lower price. Regarding pricing, until December 31, 2026, input costs $0.75, and output (including thinking tokens) costs $3.75; starting January 1, 2027, these will increase to $1.50 and $7.50, respectively.