A naukri.com initiative
VentureBeat
1d
195
Image Credit: VentureBeat
Inside Google’s AI leap: Gemini 2.5 thinks deeper, speaks smarter and codes faster
- Google announced enhancements to its Gemini 2.5 Flash and 2.5 Pro, aiming for a 'universal AI assistant' that is personal, proactive, and powerful.
- Gemini 2.5 Pro features a one-million-token context window and introduced 'Deep Think' for considering multiple hypotheses.
- Deep Think has excelled on math and coding benchmarks like USAMO and LiveCodeBench, showcasing impressive scores.
- The model leads in coding leaderboard WebDev Arena and LMArena across all categories, demonstrating proficiency.
- Google also introduced an improved 2.5 Flash model for speed, efficiency, and cost-effectiveness, second only to 2.5 Pro.
- 2.5 Flash now uses 20 to 30% fewer tokens and will be generally available for production in early June.
- New capabilities include native audio output, text-to-speech for multiple speakers, thought summaries, and thinking budgets.
- Users can steer Gemini's tone with native audio input and benefit from affective dialogue, proactive audio, and thinking features.
- Gemini 2.5 Pro and Flash support over 24 languages with expressive text-to-speech and the ability to switch between dialects smoothly.
- Developers now have the option to control thinking budgets in Pro and Flash models, providing flexibility in token usage.
- Google adds native SDK support for Model Context Protocol (MCP) definitions in the Gemini API for easier integration with open-source tools.
Read Full Article
11 Likes
For uninterrupted reading, download the app