Gemini 3.6 Flash

Google's GA Flash model delivers near-Pro coding and agentic capability at Flash-tier speed and cost, with improved token efficiency over Gemini 3.5 Flash. It is optimized for multi-step orchestration, full-stack code refactoring, general reasoning, and multimodal workflows.

Reasoning
Vision
Audio Input
Video Input
Native PDF input
Tool Calling
Structured Output

Added Jul 21, 2026

Pricing

Auto routing · per 1M tokens

Input: $0.75
Output: $3.75
Cache read: $0.075

Compare provider prices

Specifications

Context window: 1M
Max output: 65.5K
Avg output (7d): 1K tokens; Longer than 71% of models

Benchmarks

Artificial Analysis

LMArena

Design Arena

Sourced from Artificial Analysis.

Intelligence Index

34.0

Better than 88% of models compared

Coding Index

69.2

Better than 81% of models compared

Agentic Index

29.0

Better than 71% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

53.0%

Better than 70% of models compared

Harvey LAB-AA

Legal agentic work criterion pass rate

85.1%

Better than 40% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

953 Elo

Better than 46% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1286 Elo

Better than 66% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

17.4%

Better than 63% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

80.0%

Better than 86% of models compared

MLCR-AA

Medical long-context reasoning

14.4%

Better than 44% of models compared

Reasoning

HLE

Humanity's Last Exam

40.8%

Better than 88% of models compared

CritPt

Research-level physics reasoning

10.6%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

7.1%

Better than 59% of models compared

SciCode

Python programming for scientific computing

53.4%

Better than 68% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

50.0%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

55.6%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

92.8%

Better than 95% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

80.0%

Better than 86% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

39.3%

Last updated Oct 10, 2026