Gemini 3.8 Flash is launching with an introductory API price that comes with a clear deadline. Google lists the model at $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. Its standard rates are scheduled to rise to $1.50 and $7.50 on January 1, 2027.

The price change is the most important detail for developers planning long-running projects. A team can test the model at the introductory rate, but a production budget built around that number will need to be reviewed before the end of the year. Google has published the schedule in its Gemini API documentation and Google Cloud pricing material.

The model is aimed at software engineering, agentic tasks and complex multi-step reasoning. Google describes it as the company’s most intelligent workhorse model, while the Google DeepMind model card lists support for text, images, audio and video with a context window of up to one million tokens. The model card also records a maximum text output of 64,000 tokens.

Those capabilities can make a simple token comparison misleading. A system that handles a difficult task may use more reasoning steps, tool calls or output than a short chatbot request. Google’s developer documentation says teams can adjust the reasoning effort. Lower effort can reduce latency and token use for time-sensitive work, while higher effort is intended for harder jobs that need more checking.

That control gives developers a way to manage cost, but it does not remove the need for measurement. A team should record input tokens, output tokens, tool calls and the number of retries before comparing Gemini 3.8 Flash with another model. The same prompt can produce different bills when a workflow runs longer or asks the model to verify intermediate results.

Google has also announced Gemini 3.8 Flash Cyber, a separate model for vulnerability detection and automated patching through its Fairwind program. That model is not the same product as the standard Flash API. Developers should check the terms, access route and pricing that apply to their own use case instead of treating the two names as interchangeable.

The official model card includes a limitation that matters for news, research and compliance work. Gemini 3.8 Flash has a March 2026 knowledge cutoff, and Google warns that foundation models can still produce incorrect information. A current retrieval system and human review remain necessary when an answer depends on recent facts or sensitive decisions.

For developers considering Gemini 3.8 Flash, the buying question is therefore not just whether the model is faster or more capable. It is whether the workflow can stay within the introductory rate before the schedule changes, and whether its reasoning controls keep total task cost predictable. Google’s pricing page makes the date explicit, giving teams time to benchmark the model before the higher standard rates begin.