Google has made its latest AI model, Gemini 2.5 Flash, generally available to developers and enterprises, marking a significant advancement in AI coding capabilities.
The new model, which had been in preview since April, delivers substantial improvements across multiple dimensions. According to Google DeepMind, Gemini 2.5 Flash has been enhanced for reasoning, multimodality, code generation, and long context processing, while becoming even more efficient by using 20-30% fewer tokens in evaluations.
What sets Gemini 2.5 Flash apart is its status as Google's first Flash model featuring thinking capabilities. This allows the model to reason through problems before responding, with developers having the option to control the "thinking budget" to balance quality, cost, and latency. Even with thinking turned off, the model maintains the fast speeds of previous versions while improving performance.
The coding capabilities are particularly noteworthy. While Gemini 2.5 Pro leads on coding benchmarks with an impressive ELO score of 1415 on the WebDev Arena leaderboard, 2.5 Flash has also shown significant improvements in code generation and complex reasoning. This makes it valuable for developers working on everything from web applications to data processing tasks.
Beyond coding, the model introduces several new features, including native audio output for more natural conversational experiences and text-to-speech capabilities that support multiple speakers across 24 languages. Google is also bringing Project Mariner's computer use capabilities to the Gemini API and Vertex AI, allowing the model to interact with computer interfaces.
The release comes as part of Google's broader strategy to expand its Gemini 2.5 family, which now includes Flash, Pro, and the newly introduced Flash-Lite variant. All three models maintain a 1 million token context window and multimodal input capabilities, but are optimized for different use cases and efficiency requirements.
Developers can access Gemini 2.5 Flash through Google AI Studio, while enterprises can deploy it through Vertex AI. The model is also available to everyone in the Gemini app, making advanced AI capabilities more accessible across Google's ecosystem.