Google logo

Gemini 3 Flash

Multimodal
Google

Gemini 3 Flash offers frontier-level intelligence built for speed at a fraction of the cost. Combines the reasoning quality of Gemini 3 Pro with Flash-level latency, efficiency, and cost. Features a 1 million token input context window and is optimized for agentic workflows, coding, and complex analysis.

Key Specifications

Parameters
-
Context
1.0M
Release Date
December 16, 2025
Average Score
63.0%

Timeline

Key dates in the model's history
Announcement / Last Update
December 16, 2025
Today
October 5, 2026

Technical Specifications

Parameters
-
Training Tokens
-
Knowledge Cutoff
January 1, 2025
Family
-
Capabilities
MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)
$0.50
Output (per 1M tokens)
$3.00
Max Input Tokens
1.0M
Max Output Tokens
65.5K
Supported Features
Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests
SWE-Bench Verified
Single attempt • Self-reported
78.0%

Reasoning

Logical reasoning and analysis
GPQA
Without tools • Self-reported
90.0%

Other Tests

Specialized benchmarks
AIME 2025
With code • Self-reported
100.0%
MMMLU
• Self-reported
92.0%
ARC-AGI v2
ARC Prize Verified • Self-reported
33.6%
CharXiv-R
No tools • Self-reported
80.3%
FACTS Grounding
• Self-reported
61.9%
Finance Agent v2
• Verified
42.5%
Global PIQA
• Self-reported
92.8%
Humanity's Last Exam
With search and code execution • Self-reported
43.5%
Legal Agent Benchmark
All-pass rate (Harvey held-out set) • Verified
-
LiveBench
2026-01-08, High • Verified
72.4%
LiveCodeBench Pro
Elo Rating • Self-reported
77.2%
MCP Atlas
• Self-reported
57.4%
MMMU-Pro
• Self-reported
81.2%
MRCR v2 (8-needle)
1M (pointwise) • Self-reported
22.1%
OmniDocBench 1.5
Overall Edit Distance • Self-reported
12.1%
ScreenSpot Pro
No tools unless specified • Self-reported
69.1%
SimpleQA
• Self-reported
68.7%
t2-bench
• Self-reported
90.2%
Terminal-Bench 2.0
Terminus-2 harness • Self-reported
47.6%
Toolathlon
• Self-reported
49.4%
VideoMMMU
• Self-reported
86.9%

License & Metadata

License
proprietary
Announcement Date
December 16, 2025
Last Updated
December 16, 2025

Articles about Gemini 3 Flash

Similar Models

All Models

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.