Google is making waves in the AI industry with its latest announcement. The tech giant’s Google Labs team has unveiled a major upgrade to its Gemini Pro AI tool, which powers its free chatbot. This upgrade introduces the ability to process up to 1 million tokens, providing a groundbreaking context size that surpasses existing tools by a wide margin. In fact, it is 700% more powerful than OpenAI’s flagship GPT-4 model.
According to Google, this upgrade gives Gemini Pro the longest context window of any large-scale foundation model. Previously, the largest context window available for a publicly accessible language model was 200,000 tokens. Now, Google has significantly increased this to 1 million tokens consistently. This impressive feature sets a new benchmark for computational linguistics and machine learning among large language models.
However, it’s important to note that the 1 million token context window is currently only available for testing purposes. The upcoming stable version of Gemini Pro will be capable of handling up to 128K tokens, which is still a significant improvement over the previous 32,000 token limit of Gemini 1.0.
Google’s move to upgrade Gemini Pro showcases its determination to stay ahead in the AI race. Last week, the company introduced Gemini Advanced, which emerged as a strong competitor to OpenAI’s ChatGPT Plus. Unlike Anthropic’s Claude, Google’s chatbot is multimodal, performs well in various tests, and offers features that OpenAI does not.
Gemini Advanced is, however, catching up to OpenAI’s GPT-4.5 Turbo, which already has the ability to process 128,000 tokens. Nevertheless, the enhanced capabilities of Gemini 1.5 make it highly versatile. Google stated that it can process massive amounts of information in a single go, including 1 hour of video, 11 hours of audio, codebases with over 30,000 lines of code, or over 700,000 words. In fact, the Google Labs team even successfully tested Gemini 1.5 with up to 10 million tokens.
While Gemini 1.5 showcases many strengths, there are a few limitations to be aware of. For instance, it cannot analyze PDF files, which was highlighted in a comparison between Gemini and ChatGPT. However, Google has implemented the use of Mixture of Experts technology, similar to Mistral AI’s lightweight model. This technology employs a group of smaller neural networks to improve response speed and quality.
The latest upgrade to Gemini Pro demonstrates Google’s commitment to advancing its large language models. The performance of Gemini 1.5 Pro in various benchmarks surpassed that of Gemini Ultra 1.0, indicating promising prospects for Google’s LLMs.
Although a timeline for the release of Gemini Advanced 1.5 was not provided, it is evident that Google is fortifying its position in the AI industry with its enhanced token-handling capabilities. As OpenAI continues to develop GPT-5, the competition between the two tech giants remains exciting, and users can look forward to even more advanced AI models in the future.
References:
– Decrypt: [Google Is Testing an AI Model 700X More Powerful Than OpenAI’s Flagship Chatbot](https://decrypt.co/149051/google-is-testing-an-ai-model-700x-more-powerful-than-openais-flagship-chatbot)
– Google: [Google is introducing token-level capabilities in-between predictions and models](https://console.cloud.google.com/marketplace/details/google-samples/gpt-3.5-turbo-batch-assistant)
– Google Blog: [Google is introducing Gemini Advanced](https://blog.google/products.gmail/gemini-pro-gemini-advanced/)
– OpenAI: [OpenAI is developing GPT-5](https://openai.com/research/gpt-5)
