Gemini 3 Pro
Native Omni-Modal Frontier
Google DeepMind's late-2025 frontier model — handles text, images, video, and audio in one architecture (no separate "vision mode"). Comes with a Deep Think reasoning mode for hard problems and integrates across Gemini app, Vertex AI, AI Studio, and Search. Gemini 3.1 Pro / Flash / Flash-Lite refreshed the line in Q1 2026.
Intelligence
Good
Speed
Medium
122 tok/s output
Cost
High
$2.00 in / $12.00 out
Context
1M
Up to 1,000,000 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on
Artificial Analysis. Cost buckets are fixed dollar
thresholds keyed off output-token price ($/M out).
Intelligence
- Top 1%≤ 1%
- Top 5%≤ 5%
- Top 10%≤ 10%
- Good≤ 25%
- Medium≤ 50%
- Below avg> 50%
Speed
- Top 1%≥ 345 tok/s
- Top 5%≥ 237 tok/s
- Top 10%≥ 196 tok/s
- Good≥ 146 tok/s
- Medium≥ 90 tok/s
- Slow< 90 tok/s
Cost
- Freeopen weights · self-host
- Low< $1 / M out
- Moderate$1–5 / M out
- High≥ $5 / M out
Why it matters
Closed the gap with GPT-5 and Claude on reasoning while shipping natively multimodal. Established that omni-modal is now table stakes at the frontier — single-modality models look antique by comparison.
Core Capabilities
Long Documents
Handles entire codebases, books, and multi-doc RAG.
Multimodal
Combines text, vision, and audio in one model.
Generative
Produces images, video, audio, or other media.
Reasoning
Solves complex math, logic, and planning tasks.
Context Window
1M tokens
≈ entire codebase
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M This model 本模型
10M
Availability
API
Available
Product / App
Available
Open Source
Not released
Enterprise
Available
Pricing Model
Pay per token
Input and output billed separately.
Pay-per-token Capability / Performance
Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).
Lower 20% Upper 80% This model
Context / memory
Context window size · log-scaled
6.0
9.0
9.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this
Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.
What it feels like
- Tops LMArena at 1501 Elo at release — first model past 1500 on the human-preference leaderboard
- 37.5% on Humanity's Last Exam without tools, 91.9% on GPQA Diamond — PhD-level reasoning headline
- MathArena Apex 23.4% — new state-of-the-art on contest-grade math
- Multimodal: 81% MMMU-Pro, 87.6% Video-MMMU — vision and video are first-class, not bolted on
- Deep Think mode pushes GPQA Diamond to 93.8% and ARC-AGI-2 to 45.1% — biggest reasoning unlock of any 2025 model
- Reception is real but uneven — ~75% positive vs 25% negative on launch; complaints about implementation immaturity
Best use cases
- Multimodal reasoning over images, video, and long documents
- Hardest math and science research tasks (esp. with Deep Think enabled)
- Whole-codebase or long-context analysis at frontier quality
- Replacing Gemini 2.5 Pro in production once stability concerns are resolved
Tools to try
Not ideal for
- Self-hosted / open-weights deployments
- Workflows where the early-launch instability is unacceptable (some users report regressions)
- Pure SWE-bench coding leaderboards — Claude Opus 4.6 still slightly ahead