Fast, cost-effective Gemini 3 family model (preview) delivering Pro-grade reasoning at Flash-level latency and lower cost; rivals larger frontier models on PhD-level reasoning benchmarks.
profile normalized against the 47-model field
Context window· 1.05M of 1.05M100%
Max output· 66K of 272K24%
Output speed—
Affordability· $3.00 / Mtok out99%
Capability breadth· 3 of 1127%
Capability switches · 3 of 11
Reasoning mode
Tool / function use
Streaming
JSON mode
Structured outputs
Prompt caching
Fine-tuning
Web search
Code execution
Vision input
Audio input
Specifications
Every value carries a primary source and a verification date.
Capacity
Context window
1.05M
Max output
66K
Pricing
Input $/Mtok
$0.50 / Mtok input (text/image/video) USD/Mtok
Output $/Mtok
$3.00 / Mtok output USD/Mtok
Capabilities
Input modalities
["audio","image","pdf","text","video"]
Output modalities
Text
Reasoning mode
Thinking mode supported
Tool / function use
Function calling supported
Vision input
Image input supported
API
API model ID
gemini-3-flash-preview
Batch API
true
General
Knowledge cutoff
January 2025
Benchmarks
Sourced evaluation scores, each verified against its primary source.
SWE-bench Verified
On SWE-bench Verified, a benchmark for evaluating coding agent capabilities, Gemini 3 Flash achieves a score of 78%
78 %Verified
GPQA Diamond
It delivers frontier performance on PhD-level reasoning and knowledge benchmarks like GPQA Diamond (90.4%)
90.4 %Verified
Humanity's Last Exam (no tools)
and Humanity's Last Exam (33.7% without tools)
33.7 %Verified
Humanity's Last Exam (with tools)
Gemini 3 Flash: 0.435 (43.5%)
43.5 %Unverified
MMMU-Pro
It also reaches state-of-the-art performance with an impressive score of 81.2% on MMMU Pro