Google DeepMind has made available Gemini 2.0 Flash and introduced new models, including Gemini 2.0 Pro Experimental and Gemini 2.0 Flash-Lite. These updates improve performance, cost efficiency, and versatility for developers and users.
Gemini 2.0 Flash, first introduced at I/O 2024, is now widely accessible via the Gemini API in Google AI Studio and Vertex AI. The model, known for its 1 million token context window, is optimised for high-volume tasks and multimodal reasoning.
“We’ve been thrilled to see its reception by the developer community,” said Koray Kavukcuoglu, CTO of Google DeepMind. The update also brings improved benchmarks, with image generation and text-to-speech capabilities expected soon.
Alongside Gemini 2.0 Flash, Google has launched Gemini 2.0 Pro Experimental, designed for coding performance and complex prompts. This model features a 2 million token context window, the largest in the Gemini lineup, and supports tools like Google Search and code execution. “It has the strongest coding performance and ability to handle complex prompts,” Kavukcuoglu said.
The experimental version is available in Google AI Studio, Vertex AI, and the Gemini app for advanced users.
For cost-sensitive applications, Google introduced Gemini 2.0 Flash-Lite, a more efficient model that outperforms its predecessor, 1.5 Flash, on most benchmarks.
Priced competitively, it can generate captions for 40,000 photos at a cost of less than $1 in Google AI Studio’s paid tier.
Google emphasised its commitment to safety, using reinforcement learning techniques and automated red teaming to address risks, including indirect prompt injection attacks. “We’ll continue to invest in robust measures that enable safe and secure use,” Kavukcuoglu stated.
The 2.0 Flash Thinking Experimental model will also be available to Gemini app users, accessible via the model dropdown on both desktop and mobile devices.
Pricing details and further information are available on the Google for Developers blog. In the coming months, the company plans to expand multimodal capabilities and improve the Gemini 2.0 family.
Google recently updated its AI principles, removing explicit bans on using AI for weapons and surveillance. This change departs from its previous stance, which explicitly prohibited such uses since 2018.
The updated policy now focuses on risk-benefit analysis, stating that AI development will proceed where “the overall likely benefits substantially exceed the foreseeable risks and downsides.”
