Google launched experimental Gemini 1.5 models on AI Studio
New smaller Flash-8B and improved Pro models now free for developers.
Google has unveiled three new experimental Gemini 1.5 models on AI Studio, targeting developers with cutting-edge capabilities. The lineup includes Gemini 1.5 Flash-8B, an 8-billion-parameter variant optimized for multimodal applications and long-context summarization. Alongside it, a stronger Gemini 1.5 Pro model delivers significant improvements in coding and complex prompts, while the updated Gemini 1.5 Flash shows substantial performance gains across internal benchmarks. All three models boast a context window of up to 10 million tokens—far exceeding rivals like Claude 3.5 and GPT-4o—enabling processing of vast information from text, images, video, and audio. They also support function calling, allowing AI agents to interact with external tools and APIs.
These models are available for free via Google AI Studio and the Gemini API under experimental names (e.g., gemini-1.5-flash-8b-exp-0827). Early community feedback is encouraging: the new Flash model surged from #23 to #6 in the Chatbot Arena rankings based on over 20,000 votes, while Pro showed strong gains in coding and math. Google encourages developers to test these models and provide feedback, with a production-grade release expected in the coming weeks. This move positions Google to iteratively improve Gemini based on real-world usage before wider deployment.
- Gemini 1.5 Flash-8B is a compact 8B-parameter model designed for multimodal tasks and long-context summarization.
- Gemini 1.5 Pro excels at coding and complex prompts; Flash model jumped from #23 to #6 in Chatbot Arena rankings.
- All models support up to 10M token context, multimodal reasoning, and function calling, available free via AI Studio and API.
Why It Matters
Developers get free, powerful experimental models with unmatched context length to build next-gen AI applications.