On May 20, Google DeepMind began providing "Gemini 3.5," a latest suite of models that combines frontier-level intelligence with action. As the first release, Gemini 3.5 Flash has been made available to billions of users worldwide. The model outperformed Gemini 3.1 Pro in coding and agentic benchmarks (76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA, and 83.6% on MCP Atlas) and ranked first in multimodal understanding (84.2% on CharXiv Reasoning). Output speed is reported to be four times faster than other frontier models, with costs said to be less than half. The company stated that it is already using 3.5 Pro internally and plans to deploy it next month.
Sources: Gemini 3.5 Flash: frontier intelligence with action (HN 180pt, 1 comment) (HN Search (backfill))