Last December, we unveiled Gemini 1. It did not work and the Gemini API was made available in Google AI Studio and Vertex AI for enterprise customers. In February we launched Gemini 1.5 Pro, the first model with a 1 million token context window is a pro. The other month, we introduced new features that are capable of understanding native audio, system instructions, JSON mode, and so on.
Besides, we started the production of Gemma, our family of lightweight open models that are built from the same research and technology that was used to create the Gemini models. Since February, the 2B and 7B models have been downloaded millions of times, and we added the variants last month for developers and researchers, for example, CodeGemma and RecurrentGemma.
now Gemini and Gemma are becoming even more. Today, we have the task of updating Gemini 1.5 Pro, introducing 1.5 Flash announcing the unveiling of new developer features, and accession of two new Gemma models.
Gemini 1. 5 Pro additions and a new 1. 5 Flash model
Gemini 1.5 Pro: We carried out a range of improvements in the core areas of usage such as translation, coding, reasoning, and the rest. This is where you will witness these updates in the model starting today, which should assist you in solving even more complicated and wide-ranging tasks.
Gemini 1.5 Flash: The reduced Gemini model is designed for narrow or high-frequency tasks where the speed of the model’s response time is the foremost thing.
Availability: Both models are in preview now in more than 200 countries and territories and will be available generally in June.
Natively multimodal with long context: Both issue 1 and the significance of these works 1 have been discussed. 5 Pro and 1. 5 Flash have a 1 million token context window and thus you can interleave text, images, audio, and video as inputs. To get the benefits of 1, it is necessary to submit the form. 5 Pro has a 2 million token context window, therefore, you can join the waitlist in Google AI Studio or Vertex AI for Google Cloud customers.
The Gemini API’s price options and new developer features
- New developer features: According to your suggestions, we are now introducing two new API features: video frame extraction and parallel function calling, which enable you to return more than one function call at a time. Finally, in June, we will implement context caching in Gemini 1. Thus, you only need to send the parts of your prompt, including large files, to the model once. This should make the long context even more useful and even more affordable.
Pricing: Moreover, we are also increasing the rate limits supported by our new pay-as-you-go service and providing access to the Gemini API for free in eligible regions through Google AI Studio. Look at the most recent prices for Google AI Studio and Vertex AI.
Enhancements to the Gemma family
PaliGemma: The first vision-language open model is now available and it is optimized for image captioning, visual Q&A, and other image labeling tasks. PaliGemma is combined with the other pre-trained Gemma variants, CodeGemma and RecurrentGemma.
Gemma 2: The next version of Gemma will be released in June and it is designed for industry-leading performance at the most useful developer sizes. Developers and researchers have demanded a bigger model that is still easy to use. The new Gemma 27B model should do that: It is more efficient than some models that are more than twice its size and will run on GPUs or a single TPU host in Vertex AI.
