Gemini 3.8 Flash: What’s New, What It Costs, and Who It’s For

Q: What does Gemini 3.8 Flash support?

The model accepts text, images, video, audio, and PDFs and produces text. It has a 1,048,576-token input limit and a 65,536-token output limit. That input capacity can hold large document sets or code repositories, though the quality of a result still depends on how well the task and source material are organized.

Q: Standard, Batch, or Flex: which one fits?

| Mode | How it works | Best fit | Paid input / output per million tokens through 2026 | | --- | --- | --- | --- | | Standard | Immediate request-response | Interactive features | $0.75 / $3.75 | | Flex | Synchronous request with variable latency | Background steps that depend on the previous step | $0.375 / $1.875 | | Batch | Asynchronous job with a 24-hour target | Many independent, non-urgent tasks | $0.375 / $1.875 |

September 26, 2026

This post may contain affiliate links. If you buy through them we may earn a commission at no extra cost to you. Learn more.

Google’s latest Flash model is Gemini 3.8 Flash, available under the stable API ID gemini-3.8-flash. Google positions it for long-running software work, autonomous agents, and complex workflows while retaining the Flash line’s emphasis on speed and cost. The practical story is broader than raw intelligence: its large input window, multimodal inputs, tool support, and two discounted processing options make it useful in very different workflows.

What does Gemini 3.8 Flash support?

The model accepts text, images, video, audio, and PDFs and produces text. It has a 1,048,576-token input limit and a 65,536-token output limit. That input capacity can hold large document sets or code repositories, though the quality of a result still depends on how well the task and source material are organized.

Google lists Search grounding, URL context, file search, function calling, structured outputs, code execution, and context caching among its supported capabilities. Its thinking levels are low, medium, and high. Computer use is marked as a preview capability.

These details come from Google’s Gemini 3.8 Flash model page. The stable model ID matters when configuring a workflow: avoid tying an editorial system to a transient preview name when a stable version is available.

Standard, Batch, or Flex: which one fits?

Mode	How it works	Best fit	Paid input / output per million tokens through 2026
Standard	Immediate request-response	Interactive features	$0.75 / $3.75
Flex	Synchronous request with variable latency	Background steps that depend on the previous step	$0.375 / $1.875
Batch	Asynchronous job with a 24-hour target	Many independent, non-urgent tasks	$0.375 / $1.875

Google’s pricing page lists introductory rates through December 31, 2026, with higher rates starting January 1, 2027. Batch and Flex both halve the Standard token price. Flex is designed for sequential work and can take minutes; Batch is best when jobs can be submitted and collected later. Cached input has its own discounted rate, so repeated source material and editorial instructions can cost less when reused.

Can it use Google Search for current information?

Yes. The model supports Grounding with Google Search. This matters for release coverage because a model’s built-in knowledge alone cannot establish current product details. Google’s Batch API documentation also shows a batch request with the Google Search tool enabled.

Search grounding is billed separately on the paid tier after a shared allowance of 5,000 search requests per month across Gemini 3.x models; Google lists $14 per 1,000 beyond that. The free Standard tier does not include Search grounding. A single prompt can trigger multiple searches, so research workflows should track actual tool use rather than count prompts.

Who should use Gemini 3.8 Flash?

For a team processing a large volume of documents or producing structured drafts, Gemini 3.8 Flash combines a substantial context window with lower-cost Batch and Flex options. Batch fits independent jobs such as summarizing a set of release documents. Flex fits a sequence where one step feeds the next. For a reader choosing a model for everyday work, the right comparison is task-specific: use our side-by-side model comparison for current specifications and published benchmark results.

Frequently asked questions

Is Gemini 3.8 Flash free through the API?

Google lists free input and output tokens for the Standard tier, subject to the free tier’s limits. Batch and Flex do not have a free tier for this model. Search grounding is unavailable on the free tier. See the current pricing table before planning a production workflow.

Does the Batch API support Google Search grounding?

Yes. Google’s Batch API guide includes an inline request configured with google_search. Batch results are asynchronous; the target turnaround is within 24 hours.

What is the difference between Flex and Batch?

Both offer half-price token processing. Flex returns a response to a synchronous call, with variable latency and best-effort availability. Batch submits an asynchronous job and collects results later. Use Flex when the next editorial step needs the previous result; use Batch when many requests can run independently.