Gemini 1.5 Pro updates, 1.5 Flash debut and 2 new Gemma models

Grab an API key in Google AI Studio, and get started with the Gemini API Cookbook.

Last December, we released Gemini 1.0 and made the Gemini API available in Google AI Studio and Vertex AI for enterprise customers. In February, we released Gemini 1.5 Pro, the first model with a 1 million token context window. Last month, we added capabilities for native audio understanding, system instructions, JSON mode and more.

In parallel, we launched Gemma, our family of lightweight open models built from the same research and technology used to create the Gemini models. Since February, the 2B and 7B models have been downloaded millions of times, and we added variants last month for developers and researchers, like CodeGemma and RecurrentGemma.

Now Gemini and Gemma are getting even better. Today, we’re updating Gemini 1.5 Pro, introducing 1.5 Flash, rolling out new developer features and adding two new Gemma models.

Here are details on all the updates.

Gemini 1.5 Pro improvements and a new 1.5 Flash model

• Gemini 1.5 Pro: We made a series of quality improvements across key use cases, such as translation, coding, reasoning and more. You’ll see these updates in the model starting today, which should help you tackle even broader and more complex tasks.
• Gemini 1.5 Flash: This smaller Gemini model is optimized for narrower or high-frequency tasks where the speed of the model’s response time matters the most.
• Availability: Both models are available today in more than 200 countries and territories in preview and will be generally available in June.
• Natively multimodal with long context: Both 1.5 Pro and 1.5 Flash come with our 1 million token context window and allow you to interleave text, images, audio and video as inputs. To get access to 1.5 Pro with a 2 million token context window, join the waitlist in Google AI Studio or in Vertex AI for Google Cloud customers.

New developer features and pricing options for the Gemini API

• New developer features: Based on your feedback, we’re introducing two new API features: video frame extraction and parallel function calling, which lets you return more than one function call at a time. And coming in June, we’ll add context caching to Gemini 1.5 Pro, so you only have to send parts of your prompt, including large files, to the model once. This should make the long context even more useful and more affordable.
• Pricing: In addition to providing access to the Gemini API free of charge in eligible regions through Google AI Studio, we’re increasing rate limits supported by our new pay-as-you-go service. See the latest prices for Google AI Studio and Vertex AI.

Additions to the Gemma family

• PaliGemma: Our first vision-language open model is available today and optimized for image captioning, visual Q&A and other image labeling tasks. PaliGemma joins our other pre-trained Gemma variants, CodeGemma and RecurrentGemma.
• Gemma 2: Our next generation of Gemma launches in June and is built for industry-leading performance at the most useful developer sizes. Developers and researchers have requested a bigger model that’s still in a size that’s easy to use. The new Gemma 27B model should do that: it outperforms some models that are more than twice its size and will run efficiently on GPUs or a single TPU host in Vertex AI.

The first Gemini API Developer Competition

Today, we’re kicking off our first-ever Gemini API Developer Competition. We want to see your most creative and impactful apps built with Gemini models.¹Submit your projects by August 12 (and buckle up for the grand prize.)