---
title: "Gemini 3 Deep Think Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for Gemini 3 Deep Think."
source: "https://rankedagi.com/models/gemini-3-deep-think"
---

# Gemini 3 Deep Think

| Field | Value |
| --- | --- |
| Organization | Google |
| License | Proprietary |
| Type | Thinking |
| Version | Latest |
| Released | 2026-02-12 |
| Last updated | 2026-05-03T23:30:49.074Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 83.8% | RankedAGI Coding Score |  |
| RankedAGI Agentic | Agents | 82.6% | RankedAGI Agentic Score |  |
| Codeforces ELO Score | coding | 3455 | Codeforces ELO Score |  |
| RankedAGI Reasoning | reasoning | 86.1% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 48.4% | Multidisciplinary Reasoning (no tools) |  |
| Humanity's Last Exam w/ Tools | reasoning | 53.4% | Multidisciplinary Reasoning (with tools) |  |
| GPQA Diamond | reasoning | 93.8% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| MMMU-Pro | imaging | 81.5% | Multimodal Understanding |  |
| RankedAGI Math | math | 70.3% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 73.3% | Overall RankedAGI score |  |
| ARC AGI 2 | reasoning | 84.6% | Abstract Reasoning Puzzles (Public) |  |
| GDPval-AA Elo | Agents | 1324 | Office Tasks (Artificial Analysis) |  |
| FrontierScience Research | Research | 23.3% | Scientific Research |  |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think/ |
| DeepMind | https://deepmind.google/models/gemini/deep-think/ |