Google releases Gemini 3

Google releases Gemini3.7Flash: comprehensive upgrade in coding performance, initial pricing reduced by 50%. On August 13, Google officially launched the new Gemini3.7Flash just three weeks after the release of its previous model Gemini3.6Flash. Google positions it as the most powerful main model for code writing and AI Agent scenarios so far, and attributes it to major breakthroughs at the algorithm level. This model has achieved significant upgrades in software engineering and logic tasks: the FrontierCode benchmark score increased from 34.4% of the previous generation to 43.6%, and the DeepSWE test score jumped from 49.0% to 65.3%. It also performed well in web development, document understanding, and business process automation, and its overall performance has surpassed Claude Sonnet5 and GPT-5.6Terra. In terms of ecological deployment and cost control, Gemini3.7Flash has been simultaneously opened through API, AI Studio and Antigravity. Its initial pricing is set at US$0.75 per million input tokens and US$3.75 per million output tokens, which is 50% lower than the initial release price of 3.6Flash (the two models are currently priced at the same level). This preferential pricing is expected to be maintained until the end of 2026. With the performance breakthroughs and cost reduction of lightweight main models in the fields of complex coding and autonomous agents, cost-effective large models are accelerating the commercialization process of large-scale applications. via AI News (author: AI Base)