Gemini 3.8 Flash

Google's fast multimodal model for agentic workloads, including coding, tool use, image understanding, PDF and document extraction, audio, and video. Its capabilities, limits, reasoning behavior, and pricing currently mirror Gemini 3.7 Flash.

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

Added Sep 2, 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): 1.5K tokens; Longer than 83% of models

Benchmarks

Artificial Analysis

LMArena

Design Arena

Sourced from Artificial Analysis.

Intelligence Index

39.8

Better than 92% of models compared

Coding Index

74.1

Better than 89% of models compared

Agentic Index

39.7

Better than 77% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

60.9%

Better than 87% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

1180 Elo

Better than 61% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1426 Elo

Better than 78% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

22.8%

Better than 79% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

84.0%

Better than 98% of models compared

Reasoning

HLE

Humanity's Last Exam

42.1%

Better than 89% of models compared

CritPt

Research-level physics reasoning

12.3%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

19.7%

Better than 73% of models compared

SciCode

Python programming for scientific computing

55.1%

Better than 78% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

53.0%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

51.9%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

93.5%

Better than 97% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

84.0%

Better than 98% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

46.3%

Last updated Oct 12, 2026