Gemini 3.5 Flash

Google's speed-focused Gemini Flash model for frontier multimodal intelligence across text, images, audio, video, PDFs, and code. Built for agentic coding, reliable tool use, structured outputs, and long-context workflows. Audio input costs $3.00 per million audio tokens.

Reasoning
Vision
Audio Input
Video Input
Tool Calling
Structured Output

Added May 19, 2026

Pricing

Auto routing · per 1M tokens

Input: $1.50
Output: $9.00
Cache read: $0.15

Compare provider prices

Specifications

Context window: 1M
Max output: 65.5K
Avg output (7d): 3K tokens; Longer than 96% of models

Benchmarks

Artificial Analysis

LMArena

Design Arena

Sourced from Artificial Analysis.

Intelligence Index

32.6

Better than 87% of models compared

Coding Index

70.1

Better than 82% of models compared

Agentic Index

26.0

Better than 68% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

42.1%

Better than 66% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

870 Elo

Better than 47% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1259 Elo

Better than 69% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

19.8%

Better than 75% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

73.3%

Better than 74% of models compared

Reasoning

HLE

Humanity's Last Exam

42.7%

Better than 91% of models compared

IFBench

Instruction-following benchmark

76.3%

Better than 95% of models compared

CritPt

Research-level physics reasoning

13.1%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

6.6%

Better than 64% of models compared

SciCode

Python programming for scientific computing

53.9%

Better than 70% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

51.4%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

62.2%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

92.2%

Better than 94% of models compared

Terminal-Bench Hard (legacy)

Agentic coding and terminal use

40.9%

Better than 87% of models compared

T²-Bench Telecom (legacy)

Conversational AI agents in dual-control scenarios

95.3%

Better than 94% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

73.3%

Better than 69% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

35.2%

Last updated Oct 8, 2026