gemini-3.8-flash
Gemini 3.8 Flash is our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows
Modalities
In / Out Price
40% off$0.45/ 1M
$0.75 Official price
$2.25/ 1M
$3.75 Official price
Context
1.048576M
Released
Sep 2, 2026
API call examples
POST /v1/chat/completionscurl --request POST 'https://api.tokatlas.com/v1/chat/completions' \
--header "Authorization: Bearer $UNIROUTER_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "gemini-3.8-flash",
"messages": [
{
"role": "user",
"content": "Explain quantum entanglement."
}
],
"temperature": 0.7
}'Providers & routes
Current provider · Google
Platform configured price · Group: default
USD / 1M tokens. Each row shows input and output from the same tier. Displayed rates use the lowest available group multiplier; your API key group may differ.
Gemini 3.8 Flash for media analysis and multi-file work
Use it to analyze text, audio, and video in one workflow. Set boundaries for long media and cite timestamps as evidence.
Long multimodal context
Include text, images, or media with a label for each source.
Codebase tasks
Map proposed changes to affected modules and acceptance tests.
Adjustable thinking
Evaluate short questions and cross-source analysis at different levels.
Verifiable output
Include timestamps, sources, or file locations so findings can be checked individually.
About
Media and long specifications
3.8 Flash can use text, images, video, audio, and PDFs for text analysis. Its listed limits are 1,048,576 input tokens and 65,536 output tokens. Keep segment timestamps and traceable sources in media summaries.
Comparing with 3.7 Flash
Evaluate an upgrade on real multi-file coding, document, and media cases. Both versions offer large contexts. Compare visual localization, tool loops, and structured results using the same preprocessing and prompts.
Thinking, tools, and rollout
Thinking has low, medium, and high levels. Upstream features listed include function calls, code execution, file search, and grounding options. Check endpoint coverage and parsing before evaluating accuracy, cost, and timing in stages.
With long video, measure whether events remain correctly localized at different frame sampling settings. For code and documents, attach conclusions to files and sections. When migrating with grounding, tools, or caching, hold them fixed so source differences do not distort the comparison.
gemini-3.8-flash FAQ
Can it analyze recordings and documents?
It can analyze supported video, audio, images and documents into text. Confirm the media types accepted by your chosen endpoint.
How should I evaluate it in a longer workflow?
Test a task that calls tools repeatedly. Track retries and completion time before increasing the reasoning setting.
Are Gemini 3.8 Flash and Flash Cyber interchangeable?
They are separate models. Use the ID and request example shown for general-purpose Flash instead of carrying over capabilities from the other variant.
How do I pick a thinking level for Gemini 3.8 Flash?
Try a low supported setting on straightforward classification. Raise it on multi-step reasoning or broad code analysis where tests fail, and record the latency change.