Model

Gemini 2.5 Pro

CLOSED

Google · Gemini 2.5 family · released Mar 25, 2025

First Gemini with always-on thinking; #1 on LMArena at launch; official deprecation announced with successor gemini-3.1-pro-preview, shutdown scheduled ~Oct 2026.

ReasoningCodingVisionFunction callingTool useAgentic

1865.2

Elo · rank #74

Parameters

Undisclosed

Active params

Undisclosed

Context

1M tokens

Architecture

Sparse Mixture-of-Experts Transformer with native always-on thinking

License

Proprietary (Google API/Vertex AI)

Languages

—

API price (in/out)

$1.25 / $10

Modalities

text · vision · audio · video

Benchmark results

Bar shows position within the tracked field; marker = field best

Aider PolyglotCoding74.0%#10

best: Claude Opus 4.5 · 89.4%

AIMEMath86.7%#60

best: GPT-5.2 · 100.0%

GPQA DiamondReasoning84.0%#60

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning18.8%#64

best: Claude Sonnet 5 · 57.4%

LiveCodeBenchCoding70.4%#69

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MMLUKnowledge89.2%#16

best: OpenAI o3 · 92.9%

MMMUVision81.7%#18

best: Claude Fable 5 · 89.3%

PutnamBenchMath0.5%#2

best: DeepSeek-Prover-V2 671B · 7.1%

SimpleQAKnowledge52.9%#4

best: GPT-4.5 · 62.5%

SWE-bench VerifiedCoding63.8%#70

best: Claude Fable 5 · 95.0%

Video-MMEVision84.3%#9

best: Seed 2.1 Pro · 89.2%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$1.25

Output / M tok

$10

Gemini 2.5 family

Elo progression across releases

Gemini 2.5 ProMar 20251865.2 Gemini 2.5 FlashMay 20251637.3 Gemini 2.5 Flash-Lite (Non-thinking)Jun 20251193.2 Gemini 2.5 Flash-Lite (Thinking)Jun 20251281.7

Resources

DOCSai.google.dev/gemini-api/docs/models APIAvailable via Google API

API price $1.25/$10 · each benchmark row carries its own source badge (see methodology)