Google Releases Gemini 1.5 Models with Major Updates

Google releases Gemini 1.5 models with enhanced performance, reduced costs, and improved capabilities for developers. OpenAI launches an update for ChatGPT users.

Google has officially launched its Gemini 1.5 API models, bringing significant upgrades through improved performance and reduced costs. The new models, Gemini 1.5 Pro, and Gemini 1.5 Flash, offer better capabilities in various tasks like code generation, math, reasoning, and video analysis.

These models mark a leap from the earlier versions, as they display enhanced accuracy in factual responses and have reduced model hallucinations. Developers working with large datasets will also benefit from multilingual understanding across 102 languages, boosting the Gemini models’ ability to handle diverse tasks efficiently.

Price Reduction and Enhanced Performance for Gemini 1.5 Pro

Google reduced the price of the Gemini 1.5 Pro by over 50%, making it more accessible to developers while maintaining high performance. This reduction comes with better rate limits—three times higher than the older experimental versions—while offering lower latency.

Hence, developers can run their applications faster and at a fraction of the previous cost. Beginning from Oct. 1, the price for prompts with fewer than 128,000 tokens will drop.

Google outlined that input tokens would cost 64% less, while output tokens would see a 52% price cut. These changes are expected to make building applications with Gemini more affordable.

Consequently, both Gemini 1.5 Pro and Flash have improved capabilities in handling SQL generation, document analysis, and audio processing. Google has also shortened summarization lengths for both models, catering to developers working on chat-based products by offering them options to expand conversational capabilities.

Gemini 1.5 Flash-8B: An Experimental Version

In a move to support higher-volume applications, Google has increased rate limits for its models. Gemini 1.5 Flash now allows 2,000 requests per minute (RPM), while Gemini 1.5 Pro can handle up to 1,000 RPM, up from the previous 360 RPM limit.

These upgrades will make it easier for developers to scale their applications without sacrificing performance. Meanwhile, Google also launched an experimental version, Gemini 1.5 Flash-8B, which is smaller and aimed at developers needing lighter, more flexible models.

Although its benchmarks are lower than the Pro and Flash models, it still provides significant performance boosts, particularly in text and multimodal applications. These new models are available on Google AI Studio and the Gemini API.

This release underscores Google’s commitment to offering developers powerful AI tools that are both affordable and highly functional.

OpenAI Launches Advanced Voice Feature for ChatGPT Users

Meanwhile, OpenAI, a major competitor in the AI space, has rolled out its “Advanced Voice” feature for select ChatGPT users, intensifying the competition between the two AI giants. OpenAI has begun rolling out its much-anticipated “Advanced Voice” feature for ChatGPT, which is aimed at enhancing user interactions.

The update will be available to users in the ChatGPT Plus and Team tiers in phases. The new feature introduces five voices— Maple, Arbor, Spruce, SXol, and Vale— to offer a more natural and interactive experience.

These voices are an addition to the previously existing options of Juniper, Cove, Breeze, and Ember. OpenAI has focused on creating an environment where users can have smoother conversations, even allowing them to change topics mid-dialogue without interruptions.

The advanced voices will also remember past conversations, making future interactions more personalized.

ChatGPT Advanced Voice Feature Supports 50+ Languages

OpenAI also revealed that the Advanced Voice feature will support over 50 languages. This feature allows users to switch seamlessly between languages and conversations, making it more versatile for a global audience.

However, there have been some challenges with the new feature. OpenAI acknowledged in its FAQ that the voice function might not perform well with in-car Bluetooth or speakerphone devices.

Thus, there could be interruptions or issues with background noise during conversations. The rollout of these new voices comes after a delay, as the feature was originally expected earlier in the year.

However, OpenAI humorously addressed this delay by saying the voices can even say “Sorry I’m late” in multiple languages. While many users are excited about the new voices, some have drawn comparisons to a once-planned voice model called “Sky.”

However, this voice was scrapped after legal issues with actress Scarlett Johansson. Johansson claimed that OpenAI’s Sky voice closely resembled her voice, leading to legal disputes. OpenAI CEO Sam Altman denied any deliberate imitation but has since removed the voice from future updates.