Google has released Gemini 3, a new flagship AI model offering significant improvements in speed and reasoning capability. The model is now the default version in Google’s Gemini app, replacing Gemini 2.5 Flash, which Google developed for high-throughput enterprise tasks.
Gemini 3 focuses on what Google calls “agentic” tasks — the kind of multi-step problem-solving where AI must plan, decide, and act across sequential steps. The model performs well on coding tasks, a benchmark where AI performance has become a competitive differentiator among major labs.
The speed improvements are substantial. Gemini 3 processes queries faster than Gemini 2.5 Flash while delivering superior reasoning. For users accustomed to waiting for AI responses, the speed gains represent a quality-of-life upgrade. For enterprise customers running high-volume systems, faster inference means lower costs.
Google positioned Gemini 3 as the next generation in its model lineup, representing “next generation intelligence at lightning speed.” The company is updating its apps, APIs, and cloud services to use Gemini 3 as the default offering going forward.
Gemini 3.5 Flash, a variant released earlier, is engineered for sustained frontier performance on challenging agentic and coding tasks. The model hierarchy now gives enterprise customers and developers multiple options optimized for different workloads.
The release comes as competition in large language models remains intense. OpenAI, Anthropic, Meta, and other labs are releasing new models with increasing frequency, each claiming performance advances. Benchmarks on coding, reasoning, and planning tasks have become the informal scorecards for AI capability.
Pricing and availability details for Gemini 3 are being communicated to customers and developers through Google’s AI platforms and documentation. The company is positioning the release as part of its broader strategy to maintain leadership in the AI infrastructure market.
