---
title: "Gemini 3 Pro Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for Gemini 3 Pro."
source: "https://rankedagi.com/models/gemini-3-pro-preview"
---

# Gemini 3 Pro

| Field | Value |
| --- | --- |
| Organization | Google |
| License | Proprietary |
| Type | Thinking |
| Version | Preview |
| Released | 2025-11-18 |
| Last updated | 2026-05-28T21:32:30.135Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 74.3% | RankedAGI Coding Score |  |
| SWEBench Verified | coding | 76.2% | Agentic Coding | https://www.swebench.com/index.html |
| Terminal-Bench 2.0 | coding | 56.2% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| Terminal-Bench 2.1 | coding | 74.4% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.1 |
| RankedAGI Agentic | Agents | 62.4% | RankedAGI Agentic Score |  |
| BrowseComp | Agents | 59.2% | A benchmark for browsing agents | https://openai.com/index/browsecomp/ |
| Code Arena | coding | 1438 | Code Arena Elo Score | https://arena.ai/leaderboard/code |
| RankedAGI Reasoning | reasoning | 74.1% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 37.5% | Multidisciplinary Reasoning (no tools) |  |
| Humanity's Last Exam w/ Tools | reasoning | 45.8% | Multidisciplinary Reasoning (with tools) |  |
| GPQA Diamond | reasoning | 91.9% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| Vending Bench 2 | Agents | $5,478.16 | Benchmark for measuring AI model performance on running a business over long time horizons. Models are tasked with running a simulated vending machine business over a year and scored on their bank account balance at the end. | https://andonlabs.com/evals/vending-bench-2 |
| MMMU-Pro | imaging | 81.0% | Multimodal Understanding |  |
| RankedAGI Math | math | 72.9% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 64.6% | Overall RankedAGI score |  |
| 𝜏²-Bench Telecom | Agents | 98.0% | Agentic Tool Use |  |
| 𝜏²-Bench Retail | Agents | 85.3% | Agentic Tool Use |  |
| GDPval-AA Elo | Agents | 1190 | Office Tasks (Artificial Analysis) |  |
| MCP Atlas | Agents | 54.1% | Scaled Tool Use |  |
| Finance Agent | Agents | 44.1% | Agentic Financial Analysis |  |
| MMMLU | Uncategorized | 91.8% | Multilingual Q&A |  |
| Code DesignArena | Design | 1306 |  | https://www.designarena.ai/leaderboard/code |
| Toolathlon | Agents | 36.4% | Benchmark to assess language agents' general tool use in realistic environments. | https://toolathlon.xyz/introduction |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://blog.google/products-and-platforms/products/gemini/gemini-3/ |