Google has just announced the launch of its latest large language model (LLM), named Gemini 1.5. This language model is used to run the artificial intelligence (AI) based chatbot Gemini, which can converse like a human.
Gemini 1.5 is an advancement from Gemini 1.0, which was previously released by Google. This language model has capabilities equivalent to Gemini 1.0 Ultra, which is the highest version of Gemini 1.0.
One of the advantages of Gemini 1.5 is that it can accept input of about 128,000 tokens in a command or query. Tokens are units used to count the number of characters or words processed by the Gemini chatbot. One token can mean one character or one word in a sentence.
Not only that, Gemini 1.5 Pro, which is a focus variant of Gemini 1.5, can even process 1 million tokens in a single query. Google equates these 1 million tokens to context input equivalent to 1 hour of video, 11 hours of audio, 30,000 lines of programming code, or 700,000 words in a single query.
With these capabilities, Gemini 1.5 Pro can process more user requests than other existing LLM models, such as Gemini 1.0 Pro, GPT-4 Turbo, and Claude 2.1. Google also claims that Gemini 1.5 Pro has better analysis, categorization, and summarizing capabilities than other LLMs.
Gemini 1.5 can be used for various business needs, such as reviewing financial reports from different divisions, assessing the quality of movies from all stories and contexts, or creating creative content such as poems, stories, or programming code.
The early version (preview) of Gemini 1.5 Pro can already be tested by developers and Google cloud customers through the Gemini API on the AI Studio and Vertex AI platforms. Going forward, this language model will replace the Gemini 1.0 model that has been used until now.
Google will continue to develop the Gemini LLM, and is even testing query input with a limit of 10 million tokens, or 10 times more than Gemini 1.5 Pro.