Google Releases Gemini 2.5: New AI Model with Enhanced Reasoning Capabilities
2 minutes
Google DeepMind announced the release of Gemini 2.5, its latest artificial intelligence model. It is designed as a "thinking model" with improved reasoning capabilities for handling complex problems.
Key Developments
The first release in the 2.5 series is Gemini 2.5 Pro Experimental, which Google states has achieved the top position on the LMArena benchmark "by a significant margin." This benchmark measures human preferences for AI system outputs.
Koray Kavukcuoglu, CTO of Google DeepMind, described the new model as having "enhanced performance and improved accuracy" through its ability to reason through problems before responding.
Technical Capabilities
According to Google's announcement, Gemini 2.5 Pro demonstrates notable performance in:
- Reasoning tasks: The model reportedly leads in math and science benchmarks including GPQA and AIME 2025 without requiring techniques like majority voting.
- Coding performance: Google claims significant improvements over Gemini 2.0, with Gemini 2.5 Pro scoring 63.8% on SWE-Bench Verified with a custom agent setup.
- Context window: The model launches with a 1 million token context window, with plans to expand to 2 million tokens. This allows the model to process larger amounts of information in a single interaction.
- Multimodal capabilities: Gemini 2.5 maintains the ability to work with multiple types of content including text, audio, images, video, and code repositories.
Background on AI Reasoning
Google defines reasoning in AI systems as the ability to "analyze information, draw logical conclusions, incorporate context and nuance, and make informed decisions" beyond simple classification and prediction.
The company notes that this development builds on previous work in AI reasoning, including research on reinforcement learning and chain-of-thought prompting, as well as their earlier model Gemini 2.0 Flash Thinking.
Availability
Gemini 2.5 Pro Experimental is currently available through:
- Google AI Studio
- The Gemini app for Gemini Advanced users
Google has announced that the model will be coming to Vertex AI "soon" with pricing details expected "in the coming weeks."
Valeriia Kuka
Valeriia Kuka, Head of Content at Learn Prompting, is passionate about making AI and ML accessible. Valeriia previously grew a 60K+ follower AI-focused social media account, earning reposts from Stanford NLP, Amazon Research, Hugging Face, and AI researchers. She has also worked with AI/ML newsletters and global communities with 100K+ members and authored clear and concise explainers and historical articles.