Gemini 3.8 Flash

Available · Language, Multimodal, Reasoning

Gemini 3.8 Flash is a multimodal reasoning model in Google's Flash line with adjustable effort levels, released on 2 September 2026 and based on Gemini 3.7 Flash. Google aims it at long-horizon software engineering, autonomous agents and enterprise workflows; it was offered through the Gemini API, Google Antigravity, Stitch and Gemini Enterprise, and to Google AI Pro and Ultra subscribers. [1] [2] [3]September 2, 2026 Primary source

Timeline of Gemini 3.8 Flash →

Claims and evidence

  • Released Primary source[1]Gemini 3.8 Flash and Cyber: get started today [3]September 2, 2026 [4]Versions
  • Status Available Primary source[5]Gemini 3 models: gemini-3.8-flash [6]Models available for shorter availability periods
  • Successor of Gemini 3.7 Flash Primary source[1]first paragraph [2]Model Information: Description; Model dependencies
  • Change · Reasoning Takes extra reasoning steps and calls tools repeatedly on complex tasks, so Google says it can use more tokens at higher effort levels; 3.7 Flash stays supported.Compared with Gemini 3.7 Flash Primary source[1]
  • Change · Other Google says software engineering, agentic tasks and multi-step reasoning in specialised domains improved over 3.7 Flash.Compared with Gemini 3.7 Flash Primary source[2]
  • Change · Safety Google reports that the Gemini 3.8 models are more robust against prompt-injection attacks.Compared with Gemini 3.7 Flash Primary source[1]
  • Change · Availability Offered to Google AI Pro and Ultra subscribers in the Gemini app, AI Mode in Google Search and Gemini in Google Sheets, where 3.7 Flash reached consumers only through Gemini Spark.Compared with Gemini 3.7 Flash Primary source[1]
  • Input text, image, audio, videoPDF input is also listed. Primary source[7]Supported data types: Inputs [2]Model Information: Inputs
  • Output text Primary source[7]Supported data types: Output [2]Model Information: Outputs
  • Feature Reasoning modeEffort (thinking) levels low, medium (default) and high; minimal is not supported. Primary source[7]Capabilities: Thinking [2]Model Information: Description [4]introduction: thinking_level note
  • Feature Computer useListed as preview on the Gemini API model page. Primary source[7]Capabilities: Computer use
  • Context window 1,048,576 tokensThe model card gives a context window of up to 1M tokens; output is limited to 65,536 tokens. Primary source[7]Token limits: Input token limit [4]Token limits: Context window
  • Parameters Not disclosedNot disclosed by Google. No claim madeNo source
  • Access API, consumer appGemini API (Google AI Studio, Android Studio), Google Antigravity, Stitch and Gemini Enterprise; for Google AI Pro and Ultra subscribers in the Gemini app, AI Mode in Google Search and Gemini in Google Sheets. Primary source[1]Gemini 3.8 Flash and Cyber: get started today [2]Distribution

Lineage

Predecessors

Successors

No known successor.

Based on

Not derived from another model.

Variants and derived

None recorded.

Siblings

None recorded.

All ancestors

All descendants

None.

Variants

No variants recorded in this record.

Related AI Radar coverage

All model releases from Google on AI Radar →