Google Unveils Gemini 2.5 – A Smarter AI Model That Thinks Before Answering

Highlights

• Google unveiled Gemini 2.5, an AI reasoning model family designed to “think” before answering.
• Its multimodal Pro Experimental version is available for Gemini Advanced subscribers.
• Gemini 2.5 Pro boasts top benchmark scores, processing up to 1 million tokens at once (750,000 words).
• The model will compete with OpenAI’s “o” series and models from Anthropic, DeepSeek and xAI.

Caption – Google officially unveils the AI reasoning model family – Gemini 2.5. (Image credit – Google Deepmind)

Google has introduced Gemini 2.5, its latest AI reasoning model family designed to pause and “think” before responding to questions. The new model is part of a next-generation AI family focused on improving reasoning capabilities.

To kick things off, Google is launching Gemini 2.5 Pro Experimental, a multimodal AI model that it calls its most intelligent model yet. It will be available in Google AI Studio and the Gemini app for users subscribed to Google’s $20-a-month Gemini Advanced plan.

1/ Gemini 2.5 is here, and it’s our most intelligent AI model ever.

Our first 2.5 model, Gemini 2.5 Pro Experimental is a state-of-the-art thinking model, leading in a wide range of benchmarks – with impressive improvements in enhanced reasoning and coding and now #1 on… pic.twitter.com/mtEdRCTcgF

— Sundar Pichai (@sundarpichai) March 25, 2025

Moving forward, all new Google AI models will come with built-in reasoning abilities, aligning with a growing industry trend.

Think you know Gemini? 🤔 Think again.

Meet Gemini 2.5: our most intelligent model 💡 The first release is Pro Experimental, which is state-of-the-art across many benchmarks – meaning it can handle complex problems and give more accurate responses.

Try it now →… pic.twitter.com/bFcx0IlY24

— Google DeepMind (@GoogleDeepMind) March 25, 2025

🚨 Gemini 2.5 Pro Exp dropped and it’s now #1 across SEAL leaderboards:

🥇 Humanity’s Last Exam
🥇 VISTA (multimodal)
🥇 (tie) Tool Use
🥇 (tie) MultiChallenge (multi-turn)
🥉 (tie) Enigma (puzzles)

Congrats to @demishassabis @sundarpichai & team!

🔗 pic.twitter.com/06gnoqYBA4

— Alexandr Wang (@alexandr_wang) March 25, 2025

AI reasoning models are designed to take extra time and computing power to verify facts and analyze problems before generating answers. Since OpenAI launched the first AI reasoning model o1 in September 2024, other companies including Anthropic, DeepSeek, Google and xAI have been racing to develop their own.

These models have been particularly useful for tasks like math, coding, and autonomous AI agents, but they also come with higher computational costs.

Google has experimented with reasoning models before, releasing a “thinking” version of Gemini in December. However, Gemini 2.5 represents Google’s most serious attempt to compete with OpenAI’s “o” series.

Google claims that Gemini 2.5 Pro outperforms its previous AI models and many competitors on key benchmarks. The Code Editing Aider Polyglot test scored 68.6%, beating OpenAI, Anthropic, and DeepSeek.

The Software Development SWE-bench Verified test scored 63.8%, outperforming OpenAI’s o3-mini and DeepSeek’s R1, but trailing Anthropic’s Claude 3.7 Sonnet (70.3%). Meanwhile, the General Knowledge Humanity’s Last Exam scored 18.8%, surpassing most rival flagship models.

Gemini 2.5 Pro ships with a 1 million token context window, allowing it to process around 750,000 words in a single go, longer than the entire Lord of the Rings series. Google also announced that a future update will double this limit to 2 million tokens.

Google hasn’t yet shared API pricing for Gemini 2.5 Pro but promises more details in the coming weeks.

FAQs

Q1. What is Gemini 2.5 and its purpose?

Answer. Gemini 2.5 is Google’s latest AI reasoning model designed to “think” before answering, enhancing its ability to verify facts and analyze problems.

Q2. What features does Gemini 2.5 Pro offer?

Answer. Gemini 2.5 Pro supports a 1 million token context window (around 750,000 words), surpassing previous models, with plans to double the limit in future updates.

Q3. How does Gemini 2.5 compare to competitors?

Answer. Gemini 2.5 Pro outperforms many rival models on benchmarks like coding and general knowledge tests but trails slightly behind Anthropic’s Claude 3.7 Sonnet on some scores.

Also Read: Google AI Expands Globally: Major Updates for India and More

Also Read: Google Drive Now Supports Snapdragon-Powered Windows PCs