Gemini 3.8 Flash in Cursor: Pricing and Uses

Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition. is Google's latest speed-tier model in Cursor. Cursor's model page says it pairs strong agent-based coding performance with low per-token costs, and it points the model at high-throughput coding tasks that benefit from reasoning and a large context window.

Cursor does not define high-throughput on the page.

The page also does not rank Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition. against any other model. A claim that it beats a particular alternative on your codebase would have to come from your own testing, since the documentation makes no such comparison.

The model page is built around four stated strengths and no limitations section: the price per million tokens, the 90% cached-input discount, the 1M token window and the three effort levels. Each of those is covered in the sections below, and the documentation says nothing about benchmark scores, latency or tasks where the model is a poor fit.

Cursor lists three rates for Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition., all per million tokens. Fresh input is $0.75, output is $3.75, and cached input is $0.075, which Cursor describes as a 90% discount on the input rate.

Output costs five times the fresh input rate, so the output rate is the highest of the three.

The cached rate is the number that changes how a long session adds up. It is a tenth of the fresh input price, and Cursor ties it to repeated context such as large codebases. On those rates, 200,000 tokens of repeated context come to $0.15 at the fresh input price and $0.015 at the cached price.

Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition. is a third-party model, so it isn't part of the usage included with Cursor plans. On The X SubscriptionThe subscription Cursor is sold through: one plan for Grok, Cursor and X, with Free, Plus, Premium, Super and Ultra tiers, per-seat Teams, a weekly included pool for Cursor models shared with Grok, and third-party models billed at API list price from prepaid credit. Press Enter for the full definition. it is billed at these rates from your prepaid credit. On legacy Cursor plans it draws from the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool. The documentation does not state how large that pool is, so check your plan page before estimating a monthly cost.

Cursor's models and pricing page adds the plan context. Third-party models like Claude, GPT and Gemini are not part of the included pool, and on Premium and above they are available at the model's API list price from the first request, paid from prepaid credit. Plus and Free do not include third-party models. Top-ups add prepaid credit, Auto Top-up refills it when the balance drops below a threshold you set, and a monthly additional usage limit caps how much paid additional usage you can consume each month.

Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition. has three effort levels: high, medium and low. High is the default. Cursor's advice is to drop to medium or low for faster responses on simpler tasks.

Because the lower levels are the faster ones, high is the slowest of the three, and it is also where you start unless you change it. The page names faster responses as the payoff and says nothing about the price changing with the level. The page presents effort as a speed lever.

Cursor frames the lower levels as the choice for simpler tasks.

The page puts the effort advice in terms of task difficulty: high is the default, and medium or low are for faster responses on simpler tasks. It does not give a speed figure, a cost difference or a benchmark score for any of the three levels, so the only documented effect of lowering effort is faster responses.

Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition. has a 1M token context window, which Cursor says fits substantial portions of a repository in a single request. It also has access to all agent tools when used with Cursor.

Cursor ties the cached-input discount to repeated context such as large codebases.

Some plain arithmetic on the published rates shows the scale. A full 1M-token prompt costs $0.75 at the fresh input rate and $0.075 if all of it is served as cached input, and 10,000 output tokens cost $0.0375. These figures are multiplication on the listed rates, not numbers Cursor quotes, and real sessions mix fresh, cached and output tokens in proportions the documentation does not give. The models and pricing table lists no cache-write rate for Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition..

Tool access means the model is not limited to answering from what you paste into the chat. The model page links two references for the background, one on how tools work in the agent and one on tool calling fundamentals. This page does not repeat a tool list that the source only partly gives.

Choose it when the task matches what Cursor says it is for: high-throughput coding that benefits from reasoning and a large context window. Cursor describes it as a speed-tier model.

Every row above comes from Cursor's page, and the page stops there. It publishes no benchmark scores, no list of tasks where the model struggles, and no guidance on when to prefer a different model.

The page gives the prices and the effort levels, and nothing about how the model compares with others.