GoogleThe company’s latest artificial intelligence models could accelerate the adoption of AI in e-commerce and retail, developers said, as the tech giant unveiled upgrades designed to attract more companies to its business. Gemini Platform.
The company unveiled two production-ready facelift models on Tuesday (September 24). Blog PostGemini-1.5-Pro-002 and Gemini-1.5-Flash-002 offer enhanced capabilities across a range of tasks, including product recommendations, inventory management and customer service automation.
“The new release introduces advanced capabilities for math and vision tasks.” Sujan AbrahamSenior Software Engineer at an AI company Label Box“These models are Wide range It can handle tasks such as text, code, and multimodal applications. Many “They can handle more complex inputs, like 1,000-page PDFs, large code repositories, and hour-long videos. These models are faster, better, and more cost-effective for production environments.”
Lower prices and better performance
In a move that could intensify competition in the AI market, Google is slashing the price of its Gemini-1.5-Pro model by more than 50% for both input and output at prompts of less than 128,000 tokens. The company is also doubling the rate limits on the 1.5 Flash model and tripling them on the 1.5 Pro.
“To make it easier for developers to build on Gemini, we are increasing the rate limit on our paid 1.5Flash tier to 2,000 RPM. increase “The 1.5 Pro has been bumped up to 1,000 RPM from 1,000 and 360 respectively,” the post states.
Performance has improved, with Google reporting a roughly 7% increase in scores on the MMLU-Pro benchmark, which measures general knowledge and reasoning ability. Both models also showed “significant improvements of roughly 20%” in math-related benchmarks, according to the post.
“The Gemini 1.5 series is more efficient across the board.” Jorge Argota“These models are text, code and multimodal. They are more understandable and accurate when dealing with complex math and code. This could be a game changer for e-commerce platforms looking to implement advanced AI capabilities,” the AI consultancy founder told PYMNTS.
Argota highlighted the expanded context window as a key advancement.
“The model can now handle up to 2 million tokens, which is a big improvement from the previous version,” he said. “This means it can easily handle long documents and multimedia inputs. This is a big benefit for projects that deal with large datasets and long documents.”
New Features
The update also addresses concerns about speed and efficiency.
“In addition to core improvements to our latest models, over the past few weeks we’ve reduced latency by 1.5 flashes and significantly increased output tokens per second, enabling new use cases for our most powerful models,” Google’s blog post said.
Argota said:jewelry“Personalized AI Assistant” Allow Users can create custom AI assistants for specific tasks. Image 3 model, “An advanced image generation model that produces high-quality images from text prompts,” and Gemini LivePower conversational AI interactions through voice conversations.
Price cuts and increased fee limits are expected to impact businesses.
“The 15% price reduction and increased rate limits are a big win for companies looking to introduce AI into their workflows,” Argota said. “The cost savings are make These advanced models will become more affordable, especially for start-ups and SMEs who have been hesitant to adopt them due to budget constraints.”
Google is trying to make its AI models more attractive to developers as competition increases. Within the sector Like a rival Open AI and Anthropological They are also fighting for market share.
The company hopes that these improvements will encourage more developers to build applications with Gemini, leading to broader adoption of Google’s AI technology.
“We continue to be amazed by the creative and useful applications of Gemini 1.5 Pro’s 2 million token long context window and multi-modal capabilities,” the blog post reads. “From understanding videos to processing 1,000-page PDFs, there are still plenty of new use cases to build.”
In addition to the production model, Google has released an experimental version called “Gemini-1.5-Flash-8B-Exp-0924,” which the post says includes “significant performance improvements for both text and multimodal use cases.”
The update also includes changes to the default filter settings for models, giving developers more control over content moderation.
“In the model released today, filters will not be applied by default, allowing developers to determine the best configuration for their use case,” the post said.
As the AI arms race continues, Google’s latest moves demonstrate the company’s determination to stay competitive in a rapidly evolving market. With these enhancements, the company aims to solidify its position as a go-to AI tools provider for developers and businesses.
“Overall, this is a big win for companies looking to add AI to their workflows,” Argota said.
To stay up to date on all things PYMNTS AI, subscribe to our daily AI newsletter.
