Gemini 2.5 Pro Preview 06-05

The latest preview version of Google's most advanced reasoning Gemini model, capable of solving complex problems. Built for the agentic era with enhanced reasoning capabilities, multimodal understanding (text, image, video, audio), and a 1M token context window. Features thinking preview, code execution, grounding with Google Search, system instructions, function calling, and controlled generation. Supports up to 3,000 images per prompt, 45-60 minutes of video, and 8.4 hours of audio.

Benchmark results

Benchmark Score Tags Source
Aider-Polyglot 82.2% self-reported llm-stats link →
AIME 2025 88.0% self-reported llm-stats link →
FACTS Grounding 87.8% self-reported llm-stats link →
Global-MMLU-Lite 89.2% self-reported llm-stats link →
GPQA 86.4% self-reported llm-stats link →
Humanity's Last Exam 21.6% self-reported llm-stats link →
LiveCodeBench 69.0% self-reported llm-stats link →
MMMU 82.0% self-reported llm-stats link →
MRCR v2 (8-needle) 16.4% self-reported llm-stats link →
SimpleQA 54.0% self-reported llm-stats link →
SWE-Bench Verified 67.2% self-reported llm-stats link →
Vibe-Eval 67.2% self-reported llm-stats link →
VideoMMMU 83.6% self-reported llm-stats link →