Gemini 3.8 Flash Is Here: Same Price, Stronger Reasoning, and Better PDF Handling

Google has shipped Gemini 3.8 Flash, its newest Flash-tier AI model, on September 2, 2026. It reads text, images, video, audio, and PDFs, runs on a 1,048,576-token context window, and, unlike most model upgrades, it costs exactly the same as the version it replaces. For anyone already using Gemini, this is essentially a free upgrade.

Google didn't hold a big keynote for this one, it landed through the developer changelog. But for anyone who regularly hands documents to an AI model, whether that's a PDF contract, a scanned form, or a long report, the details underneath are worth knowing.

What's Actually New in Gemini 3.8 Flash

Google describes Gemini 3.8 Flash as its most intelligent Flash model yet, built for long-horizon software engineering, autonomous agents, and complex enterprise workflows. It keeps the same modality support as its predecessor, text, image, video, audio, and PDF input, and returns text output, with the context window holding steady at just over 1 million tokens.

The gains are real but specific. Independent benchmarking from Artificial Analysis shows the biggest jump in agentic, multi-step task handling, up nearly five points over Gemini 3.7 Flash, the largest single improvement anywhere in the comparison. Coding performance, by contrast, barely moved. If your work involves the model planning and following through on longer, multi-part tasks, this update is aimed squarely at you. If you mainly use it for straight code generation, you likely won't notice much of a difference.

There's one trade-off worth knowing before you switch. Gemini 3.8 Flash is measurably slower to start responding than 3.7 Flash, even though it streams faster once it gets going. For latency-sensitive, real-time chat use, that's a small but real regression worth testing against your own workflow before committing.

Same Price, Real Upgrade, For Now

Pricing is where this release stands out. Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, identical to Gemini 3.7 Flash on the same million-token context. Batch and Flex inference cut that roughly in half, while Priority inference costs more for faster response times.

Worth flagging clearly: these are introductory rates. Google has said pricing on both tiers will roughly double starting January 1, 2027, so this is a good window to test and build on today's pricing before that changes.

Why the PDF Support Actually Matters Here

PDF input isn't a new feature for Gemini, but pairing it with a 1-million-token context window and stronger agentic reasoning changes what's realistic to do with it. A model that can hold an entire long document in memory while reasoning through multi-step instructions can genuinely work through a full contract, research paper, or lengthy report in one pass, rather than losing track of earlier sections partway through.

As with any large-context model, the quality of what goes in still shapes the quality of what comes out. A scanned PDF full of image-based pages, an oversized file, or a document that hasn't been converted into clean, readable text will eat through that context budget fast and give the model less to actually reason over. Preparing a file first, compressing it, converting it to a cleaner format, or extracting just the pages that matter, still makes a real difference before it goes anywhere near an AI model. Our PDF Compressor, PDF to Word, and document tools handle exactly that kind of prep in your browser, no installs needed.

Where This Fits Among Google's Other Recent Updates

This release lands alongside a broader push across Google's AI lineup, Gemini can already generate downloadable PDFs, Word docs, and Excel files directly inside a conversation, and Gemini in Drive, Docs, and Sheets has been steadily gaining features that pull information from a user's own files to draft documents and build spreadsheets automatically. Gemini 3.8 Flash's stronger agentic reasoning is likely to feed directly into those same document-handling features going forward.

Who Should Actually Care About This Update

If you're a developer building on the Gemini API, this is a meaningful free upgrade worth testing against your existing 3.7 Flash workflows, especially anything involving multi-step tasks or long documents. If you're just using the Gemini app day to day, the improvements will likely show up quietly in the background rather than as something you need to turn on.

We'll keep tracking how Gemini's document and PDF handling evolves, and how it compares to similar updates from ChatGPT and Adobe's Acrobat AI, in our AI tools and updates section.