Google Advanced Available in FlowHunt

Gemini 3.8 Flash

Gemini 3.8 Flash is an advanced-tier AI model from Google, released in September 2026. It is a proprietary closed-weight model, available through the Gemini app, Google AI Studio, the Gemini API, Google AI Mode, Google Antigravity, and the Gemini Enterprise Agent Platform. It accepts text, images, audio, and video, with a context window of up to 1M tokens.

Read more

Gemini 3.8 Flash was published on September 2, 2026 as the next iteration in the Gemini 3 family, building on Gemini 3.7 Flash. Google describes the gains as being in software engineering and agentic knowledge workflows. The model keeps customizable effort levels, which let you trade quality against cost and latency. It has a 1M-token context window.

Google positions it for cost-effective scaling of general-purpose, production-ready agents, with software engineering, agent tasks, and complex knowledge workflows as the main use cases.

Try in FlowHunt Back to AI Leaderboard

Benchmarks

Each score appears once, with its rank among tracked models and who measured it.

Compared with Advanced peers on the same scale
  • Gemini 3.8 Flashthis model
  • Mistral Large 4this model leads on 3 of 4
  • Claude Sonnet 5.5this model leads on 1 of 4
Verified independent evaluator, counted in the ranking
Overall

AA Intelligence Index

Independent composite of agent, coding, knowledge and science tests by Artificial Analysis (v4).

Source ↗

Gemini 3.8 Flash 40.9 Mistral Large 4 38.4 Claude Sonnet 5.5 56.0
#17of 45 Verified
Coding

DeepSWE

113 long-horizon software tasks written from scratch, graded by hidden tests.

Source ↗Also: Self-reported 73.7%

Gemini 3.8 Flash 73.8% Mistral Large 4 61.7% Claude Sonnet 5.5 71.0%
#2of 17 Verified

SWE-bench Verified

500 verified GitHub issues. Saturated at the frontier.

Source ↗

Gemini 3.8 Flash 80.0% Mistral Large 4 No score Claude Sonnet 5.5 No score
#11of 22 Verified
Agents

Terminal-Bench 4.0

Multi-step professional tasks done by an agent in a command-line terminal.

Source ↗Also: Self-reported 19.1%

Gemini 3.8 Flash 19.7% Mistral Large 4 26.8% Claude Sonnet 5.5 63.6%
#18of 36 Verified
Reasoning

Humanity's Last Exam

Expert-written questions across 100+ subjects, no tools.

Source ↗

Gemini 3.8 Flash 47.8% Mistral Large 4 35.0% Claude Sonnet 5.5 55.0%
#11of 44 Verified

GPQA Diamond

Graduate-level science questions that search cannot answer.

Source ↗Also: Verified 94.4% (Vals AI own implementation)

Gemini 3.8 Flash 95.3% Mistral Large 4 No score Claude Sonnet 5.5 No score
#2of 37 Verified

ARC-AGI-2

Novel visual puzzles that cannot be memorised.

Source ↗

Gemini 3.8 Flash 89.2% Mistral Large 4 No score Claude Sonnet 5.5 No score
#7of 27 Verified

What to Use Gemini 3.8 Flash For

Software Engineering

Google lists software engineering as a main use case and as an area of improvement over Gemini 3.7 Flash. Use it for code generation, issue resolution, and multi-step coding tasks.

Production Agents at Scale

It is designed for cost-effective scaling of general-purpose agents. Adjustable effort levels let you spend more reasoning on hard steps and less on routine ones.

Complex Knowledge Workflows

Use it for multi-step knowledge work such as research, analysis, and document-heavy business tasks.

Long-Document & Multimodal Analysis

With text, image, audio, and video input and a 1M-token context window, Gemini 3.8 Flash can work through long documents, large codebases, and mixed-media material in a single prompt.

Strengths and limitations

Strengths

  • Software engineering and agentic workflows
  • Complex knowledge workflows
  • Adjustable effort levels for quality, cost, and latency
  • Text, image, audio, and video input
  • Very long documents and large codebases (1M context)

Limitations

  • Closed-weight, so it cannot be self-hosted or fine-tuned on private data
  • Google notes it can hallucinate, may be occasionally slow, and may use more tokens at higher effort levels

Use it in FlowHunt

Use Gemini 3.8 Flash in any flow, next to your own tools and knowledge sources.

  1. 1Open a flow, or create one from a template
  2. 2Add an AI Agent or Generator component
  3. 3Pick Gemini 3.8 Flash in the model selector

Related AI Models

About Google Mountain View, CA · Founded 1998 · ai.google

Google is a technology company founded in 1998 and headquartered in Mountain View, California. Its Gemini family of models is developed by Google DeepMind, the research lab formed from Google Brain and DeepMind.

ai.google ↗

Browse the Full AI Model Leaderboard

Compare every tracked model on coding, agent and reasoning benchmarks, with each score marked verified or self-reported.

← Back to AI Leaderboard Try in FlowHunt

Frequently asked questions