Google is expanding Gemini 1.5, its most powerful family of AI models, with new variants.

On Tuesday, Google AI Studio product lead Logan Kilpatrick announced at X that the company had released three new experimental versions of Gemini: a smaller model, Gemini 1.5 Flash-8B; a “more powerful” model, Gemini 1.5 Pro; and the “significantly improved” Gemini 1.5 Flash.

Kilpatrick explained that Google is “releasing experimental models to gather feedback and get our latest updates into the hands of developers.”

Also: Google’s Gemini just got two popular ChatGPT Plus features – and one of them is free

The first model, 1.5 Flash-8B, is an eight-billion-parameter version of the new 1.5 Flash model and can be used for “everything from high-volume multimodal use cases to long-time context summarization tasks,” Kilpatrick noted in the X thread.

The new version of Pro includes improvements in math, complex prompts and coding, while the new Flash now scores better on certain internal benchmarks. Kilpatrick said Gemini 1.5 Pro Exp 0827 (as of release day, August 27) will replace the most recently released model 0801. Starting September 3, 0801 will be redirected to model 0827 in the Gemini API.

Also: Want to try GPT-4o mini? 3 ways to access the smarter, cheaper AI model – and 2 are free

Shortly after their release, the latest Gemini 1.5 Pro ranked 2nd and Flash 6th in the Chatbot Arena, putting them practically neck and neck with GPT-4o and GPT-4o mini respectively. The two models beat Claude 3.5 Sonnet, Grok 2, Grok 2 mini and Llama 3.1.

The experimental models join the Gemini 1.5 family, which is designed to handle very long context windows. In a technical report earlier this month, the DeepMind team called their capabilities “unprecedented among contemporary large-scale language models (LLMs)” and stated that Gemini 1.5 can handle multimodal inputs such as “entire document collections, several hours of video, and nearly five days of audio.”

The team added that these new versions continue the models’ trend of “near-perfect recall (>99%) up to at least 10 million tokens,” compared to Claude 3.0’s 200,000 tokens and GPT-4 Turbo’s 128,000 tokens.

The report also mentioned potential use cases and praised Gemini 1.5’s ability to help working professionals save up to 75% of their time on tasks across ten job categories. Among other things, it imparted some surprising “game-changing” capabilities: “Given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English into Kalamang at a similar level to a person who learned with the same content,” the report said.

Also: ChatGPT vs. Microsoft Copilot vs. Gemini: Which is the best AI chatbot?

However, the response to the experimental models was mixed. Some users praising Google’s fast releases, while others, unimpressedasked for the release of Gemini 2.0 instead. asked by an X user about benchmarks, Kilpatrick replied that the company plans to release a production version, hopefully including evaluations, “in the coming weeks”!

Users can test all three models for free today in Google AI Studio and the Gemini API.