Gemini 2.5 Pro is presented as Google's most capable model to date, achieving state-of-the-art (SoTA) performance on coding and reasoning benchmarks. It excels in multimodal understanding and can process up to 3 hours of video content. The model's long context, multimodal, and reasoning capabilities facilitate new agentic workflows. In comparison to its predecessors, Gemini 2.5 Pro demonstrates marked improvements in coding, math, and reasoning tasks. It has shown noticeable enhancements in image understanding as well.

The Gemini 2.5 models show significant improvements in coding tasks such as LiveCodeBench, Aider Polyglot, and SWE-bench Verified. Performance on LiveCodeBench increased from 30.5% for Gemini 1.5 Pro to 69.0% for Gemini 2.5 Pro, while performance on Aider Polyglot went from 16.9% to 82.2%. Additionally, SWE-bench Verified performance increased from 34.2% to 67.2%. Relative to other available large language models, Gemini 2.5 Pro achieves the SoTA score on the Aider Polyglot coding task.

Gemini 2.5 models are better at math and reasoning tasks than Gemini 1.5 models. For example, performance on AIME 2025 is 88.0% for Gemini 2.5 Pro compared to 17.5% for Gemini 1.5 Pro, and performance on GPQA (diamond) went from 58.1% for Gemini 1.5 Pro to 86.4%. Furthermore, Gemini achieves the highest score on GPQA (diamond) out of all the models examined.

Gemini 2.5 Pro also achieves the highest score on SimpleQA and FACTS Grounding factuality benchmarks compared to other models. These benchmarks evaluate the model's ability to provide accurate responses to information-seeking prompts and generate responses grounded in user-provided documents.

The Gemini 2.5 models showcase significant improvements in long context understanding. They achieve state-of-the-art quality on LOFT and MRCR long-context tasks at 128k context. Gemini is also the only model examined that supports context lengths of 1M+ tokens. The new 2.5 models improve greatly over previous Gemini 1.5 models and achieve state-of-the-art quality on all of those tasks.

Gemini’s multilingual capabilities have undergone a profound evolution since 1.5, which already encompassed over 400 languages via pretraining. The impact is particularly striking in Indic and Chinese, Japanese and Korean languages, where dedicated optimizations in data quality and evaluation have unlocked dramatic gains in both quality and decoding speed. Consequently, users benefit from significantly enhanced language adherence, responses designed to faithfully respect the requested output language, and a robust improvement in generative quality and factuality across languages, solidifying Gemini’s reliability across diverse linguistic contexts.

Gemini 2.5 Pro demonstrates state-of-the-art audio understanding performance as measured by public benchmarks for ASR and AST, and compares favorably to alternatives under comparable testing conditions (using the same prompts and inputs).

Gemini 2.5 Pro achieves state-of-the-art performance on key video understanding benchmarks, surpassing recent models like GPT 4.1 under comparable testing conditions (same prompt and video frames). For cost-sensitive applications, Gemini 2.5 Flash provides a highly competitive alternative.

Gemini 2.0 marked a significant leap as our first model family trained to natively call tools like Google Search, enabling it to formulate precise queries and synthesize fresh information with sources. Building on this, Gemini 2.5 integrates advanced reasoning, allowing it to interleave these search capabilities with internal thought processes to answer complex, multi-hop queries and execute long-horizon tasks. The model has learned to use search and other tools, reason about the outputs, and issue additional, detailed follow-up queries to expand the information available to it and to verify the factual accuracy of the response.

The Gemini 2.5 Flash model has become the second most capable model in the Gemini family, overtaking not just previous Flash models but also the Gemini 1.5 Pro model released one year prior.

The Gemini 2.5 models maintain robust Safety metrics while improving dramatically on helpfulness and general tone compared to their 2.0 and 1.5 counterparts. In practice, this means that the 2.5 models are substantially better at providing safe responses without interfering with important use cases or lecturing end users.

How many global developers use NVIDIA AI ecosystem?

Related Content You May Like