Google Gemini 2.5: New Abilities
- Google has launched Gemini 2.5, a new family of AI reasoning models designed to "think" before providing answers.
- Google says future AI models will feature innate reasoning capabilities.
- Google asserts that Gemini 2.5 Pro surpasses previous AI models,including those from major competitors,across various benchmarks.The model is designed for creating visually engaging web applications and agent-based programming...
Google Introduces Gemini 2.5 AI Reasoning Models
Google has launched Gemini 2.5, a new family of AI reasoning models designed to “think” before providing answers. The initial release, Gemini 2.5 Pro, is available on the Google AI Studio developer platform and within the Gemini application for Gemini Advanced subscribers, a $20 per month package.
About gemini 2.5
Google says future AI models will feature innate reasoning capabilities. The company previously experimented with reasoning-based AI, including a thinking
version of Gemini in December 2024.Gemini 2.5 represents GoogleS effort to compete with OpenAI’s O series models.
Google asserts that Gemini 2.5 Pro surpasses previous AI models,including those from major competitors,across various benchmarks.The model is designed for creating visually engaging web applications and agent-based programming applications.
In code editing evaluations using Aider Polyglot, Gemini 2.5 Pro achieved a score of 68.6%, outperforming AI models from OpenAI, Anthropic, and Deepseek. On the Swe-Bench Verified software progress capability test, the model scored 63.8%, exceeding OpenAI’s O3-Mini and Deepseek R1, but falling short of Anthropic’s Claude 3.7 Sonnet, which reached 70.3%.
On a multimodal examination assessing mathematical, humanities, and scientific knowledge, Gemini 2.5 pro scored 18.8%, exceeding most competing AI models.
Text Commands, Audio, Images, videos, and Code
According to Google DeepMind CEO Demis Hassabis, the Gemini 2.5 Pro model demonstrates a elegant improvement of +39 points, with important gains across multimodal understanding, coding, and STEM reasoning.
Gemini 2.5 Pro launched with a context window of 1 million tokens, enabling the model to process approximately 750,000 words at once. Google plans to expand this capacity to 2 million tokens. pricing for the Gemini 2.5 Pro API has not yet been announced; further details are expected in the coming weeks.
as OpenAI introduced its O1 reasoning AI model in September 2024, the tech industry has been actively developing models with similar or superior capabilities. Companies like Anthropic, Deepseek, Google, and XAI currently offer AI models with advanced reasoning.
These models utilize additional computing power to verify facts and reason through problems before providing answers, enhancing accuracy in tasks such as mathematics and programming.
Additional Features
Gemini has also launched new Canvas and Audio Overview features.
Gemini 2.5 AI: Your Essential Q&A Guide
Updated: March 26, 2025
What is Gemini 2.5?
Gemini 2.5 is a new family of AI reasoning models developed by Google, designed to “think” before providing answers. These models represent Google’s latest advancements in the field of artificial intelligence, offering improved reasoning capabilities and performance across various tasks.
What are the Key Features of Gemini 2.5 Pro?
- Enhanced Reasoning: Gemini 2.5 Pro is designed to “think” before answering, improving accuracy in complex tasks.
- Multimodal Capabilities: It supports text, audio, images, videos, and code.
- Large Context Window: Launched with a context window of 1 million tokens (approximately 750,000 words) and plans to expand to 2 million tokens.
- New Features: Includes new Canvas and Audio Overview features.
how Does Gemini 2.5 Perform?
Gemini 2.5 Pro has demonstrated strong performance across multiple benchmarks, surpassing many competing AI models. Here’s a quick look at its performance:
- Code Editing (Aider Polyglot): Achieved 68.6%, outperforming AI models from OpenAI, Anthropic, and Deepseek.
- Software Progress (Swe-Bench Verified): Scored 63.8%, exceeding OpenAI’s O3-Mini and Deepseek R1, but falling short of Anthropic’s Claude 3.7 Sonnet (70.3%).
- Multimodal Examination: Scored 18.8% on a multimodal examination assessing mathematical, humanities, and scientific knowledge, surpassing most competing AI models.
How Does Gemini 2.5 Compare to Other AI Models?
Gemini 2.5 is designed to compete directly with models like OpenAI’s O series. The AI landscape is highly competitive, with various companies offering models with advanced reasoning capabilities. Here’s a comparison based on available data
| Model | Benchmark | score | Key Competitors |
|---|---|---|---|
| Gemini 2.5 Pro | Code Editing (aider Polyglot) | 68.6% | OpenAI,Anthropic,Deepseek |
| Gemini 2.5 Pro | Swe-Bench Verified | 63.8% | openai (O3-Mini), Deepseek R1, Anthropic (Claude 3.7 Sonnet) |
| Gemini 2.5 Pro | Multimodal Examination | 18.8% | Various AI models |
Were Can I Access Gemini 2.5?
The initial release of Gemini 2.5 Pro is available on the Google AI Studio developer platform and within the Gemini application for Gemini Advanced subscribers.
What is the Cost of Gemini 2.5 Pro?
Currently, pricing details for the Gemini 2.5 Pro API have not been announced. Further data is expected to be released in the coming weeks.
What are the plans for the future?
Google plans to expand the context window to 2 million tokens. they are also expected to release future models that include innate reasoning capabilities.
Who are the main competitors of Gemini 2.5?
Key competitors include OpenAI, Anthropic, Deepseek, and XAI. All of these companies are actively developing AI models with advanced reasoning capabilities.