All benchmarks
Google DeepMind · USA · Released 2025-06
Gemini 2.5 Pro
Google's most capable model with native 1M-token context and full multimodal I/O across text, images, audio and video.
Access
closed
Context
1,000,000 tokens
Modalities
text, vision, audio, video
Languages
100+
Benchmark scores
MMLU
89.8%
Massive Multitask Language Understanding
GPQA
86.4%
Graduate-level science QA (Diamond)
HumanEval
92.6%
Python code completion
SWE-bench
63.8%
Real GitHub issue resolution (Verified)
LiveCodeBench
N/A
Contamination-free coding
AIME
88.0%
American Invitational Mathematics Exam
MMMU
82.0%
Multimodal college-level reasoning
MATH
91.5%
Competition-level math word problems
Scores as reported by the vendor or leading public leaderboards. "N/A" means the score has not been publicly disclosed for this metric.
Explain like I'm 5
Reads whole books at once and understands pictures, sound, and video together.
Key features
- 1M token context
- Native video understanding
- Deep Think mode
- Grounded search
Strengths
- Best long-context
- Strong multimodal reasoning
- Fast on Google infra
Limitations
- Occasional refusals on edge cases
Best for
Long-doc analysisVideo Q&AMultimodal agents
Related models
OpenAI
GPT-5
OpenAI's flagship reasoning + multimodal model with a unified router that switches between fast and deep-thinking modes.
Anthropic
Claude Sonnet 4.5
Anthropic's best coding + agentic model, purpose-built for long-horizon computer-use and software engineering tasks.
xAI
Grok 4
xAI's frontier reasoning model with real-time X data grounding and a heavy multi-agent variant (Grok 4 Heavy).
Meta AI
Llama 4 Maverick
Meta's flagship MoE model — 400B total / 17B active — with native multimodal input and 1M-token context.