Gemini 2.5 Pro
Google DeepMind · March 2025
● activeClosedmixture of expertsmultimodalAPI Available
Context Window1M tokens
VariantsPro, Flash
Why It Matters
Google's entry into the 'thinking model' category, competing with OpenAI's o-series and Anthropic's extended thinking. Topped the LMArena leaderboard at launch.
Description
Google's first 'thinking model' — designed to reason through complex problems by analyzing information and drawing logical conclusions before responding, similar to how a human expert would think through a difficult question. Debuted at the top of the LMArena leaderboard with major improvements in coding, math, and multi-step reasoning.
Notable Milestones
- ▸Debuted at #1 on LMArena leaderboard
- ▸Flash variant offered thinking capabilities at much lower cost
Benchmark Scores
GPQAGraduate-level science QA
84.0%AIMEAMC/AIME math competition
86.7%SWE-benchReal-world software engineering
63.8%Key Innovations
Reasoning
ReasoningStructured step-by-step problem solving, often using chain-of-thought or tree-of-thought approaches.
Agentic
AgenticModels that can autonomously plan, execute multi-step tasks, use tools, and self-correct without human intervention.
Test-Time Compute
Test-Time ComputeUsing extra computation during inference (not training) to improve answer quality — thinking longer on harder problems.
Family Tree
Built On
Lineage
Successors (1)
Related Research (1)
GeminiScaling
2023 · Google DeepMind
Introduced the Gemini family with native multimodal training from the ground up, achieving SOTA on 30+ benchmarks.