Gemini 3 Pro
Google Gemini 3 Pro: Multimodal reasoning model with dynamic thinking_level control, thought signatures, media_resolution parameters, and 1M token context.
Model Overview
Capabilities, design details, and architectural traits
Gemini 3 Pro - Natively Multimodal Reasoning with Dynamic Thinking by Default
Gemini 3 Pro is Google's Pro-tier model in the Gemini 3 series. It is natively multimodal - accepting text, images, video, audio, and PDFs as inputs to a single reasoning model. The original gemini-3-pro-preview endpoint now redirects to gemini-3.1-pro-preview, its direct successor based on the same architecture.
| Trait | Detail |
|---|---|
thinking_level parameter | Controls the maximum depth of internal reasoning before responding; replaces the legacy thinking_budget parameter; defaults to high (dynamic) for Gemini 3 Pro |
| Thought signatures | Encrypted representations of the model's internal reasoning process, maintained across API calls; required in stateless mode; automatically managed in stateful mode via the Interactions API |
media_resolution parameter | Gemini 3-specific API parameter for granular control over multimodal vision processing resolution per request |
| Context window | 1M token input, 64k token output |
How thinking_level Differs from Prior Gemini Generations
Gemini 2.x used a thinking_budget parameter. Gemini 3 replaces this with thinking_level (low, medium, high), which sets a relative reasoning depth allowance rather than a strict token count. The model treats these as guidance, not hard limits. Mixing both parameters in the same request returns a 400 error - they are mutually exclusive.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | |
| Release Date | November 18, 2025 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 1.0M Best Context Window | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $2 Best Input Pricing | $4 | $2 Best Input Pricing |
| Output Pricing | $12 | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | audiofileimagetextvideo | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 28.0 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Google
Other models by Google
Gemini 4 Argon
Gemini 3.8 Flash
Gemini 3.7 Flash
Gemini 3.6 Flash
Top AI Models
Leading alternatives by intelligence score
