Overview

• Paid subscribers can now submit up to 100 prompts per day to Gemini 2.5 Pro, up from 50.
• The update boosts coding performance to a new high of 82.2% on the Aider Polyglot test, outpacing OpenAI, Anthropic and DeepSeek.
• An experimental “Deep Think” mode has been added to enhance the model’s reasoning on complex math and coding tasks.
• Response formatting has been improved with more creative outputs and developers can configure thinking budgets to manage resource use.
• Google aims to roll out this build as the long-term stable release, signaling confidence in light of growing demand.