Gemini 3.7 Flash: Pricing, Context and Uses

Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. is Google's latest speed-tier model in Cursor, with the model ID gemini-3.7-flash. Cursor rates it fast for speed, low for cost, and high for intelligence. It draws from the third-party Other ModelsCursor's pool of supported third-party models from providers such as Anthropic, Google and OpenAI, with separate plan access and usage from Cursor Models. Press Enter for the full definition. pool rather than the Cursor ModelsCursor's plan and usage pool for models it labels as Cursor Models, kept separate from the third-party Other Models pool. Press Enter for the full definition. pool. That makes it a direct model choice for quick, reasoning-heavy coding work, not a replacement for the larger included pool attached to Grok and ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition..

The useful part of the release is the combination of price and context. Fresh input costs $0.75 per million tokens, while a cached read costs $0.075. The model card lists 200k as the standard context and 1M as the maximum. A long session can therefore keep a substantial working set available without paying a frontier-model output rate on every response.

Use Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. when the task needs more than a single mechanical edit but still rewards speed. Repository searches followed by several related changes, a test failure that needs a short investigation, or a frontend fix that must be checked in the browser all fit the shape Cursor describes. Give it a bounded outcome and a command or screen state that can prove the work is done.

High-throughput does not mean unattended. Start with the files most likely to own the behavior, ask the model to inspect before editing, and require the relevant tests afterward. The low output price helps when a task takes several tool calls, but a cheap failed attempt still creates review work. For me, the sensible default is a reviewable task that can finish inside one focused session.

Cursor lists Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. at $0.75 per million input tokens, $0.075 per million cache reads, and $3.50 per million output. The cache-write column is blank, so its rate is not published. Do not turn that missing value into a claim that cache writes are free. Usage comes from the Other ModelsCursor's pool of supported third-party models from providers such as Anthropic, Google and OpenAI, with separate plan access and usage from Cursor Models. Press Enter for the full definition. pool at the published model rate.

Plan access and usage balance are separate questions. Cursor lists Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. in the Other ModelsCursor's pool of supported third-party models from providers such as Anthropic, Google and OpenAI, with separate plan access and usage from Cursor Models. Press Enter for the full definition. pool available to Pro, Pro Plus and Ultra. Cursor StartAn India-only individual plan centered on Cursor Models, with cloud and mobile access but without Other Models, Auto, Bugbot, Automations or the Agent SDK. Press Enter for the full definition. does not include that pool. Once selected, its tokens use the third-party allowance described in Cursor pricing. Check the live picker before making a rollout depend on one model.

Teams and Enterprise add the Cursor Token Rate to eligible third-party model requests. The current fee is $0.25 per million tokens and applies to input, output, and cached tokens. A directly selected cached read therefore combines the $0.075 model rate with that separate fee. Check the usage record and contract before turning a model-table row into a team budget.

Treat 200k as the normal working window and 1M as the documented ceiling. A million-token maximum can hold a large repository slice, but it does not make every file relevant. Begin with the feature boundary, search outward when imports or test failures point elsewhere, and keep the acceptance check in view. More context is useful only while the model can still find the evidence that controls the change.

The 90% cache-read discount matters when the same context appears again. A first pass over one million fresh input tokens is $0.75 at the published model rate; a later cached read of the same amount is $0.075. That saving fits a sustained codebase session much better than a set of unrelated prompts. The blank cache-write rate also means you should validate the actual usage record instead of estimating the whole session from cache reads alone.

1. 1Open a fresh session for one task and name the expected output and verification command.
2. 2Load the smallest repository slice that explains the behavior, then search for missing dependencies.
3. 3Keep stable instructions and relevant files in the same session so repeated context can be cached.
4. 4Check the context indicator before adding broad logs, generated files, or unrelated folders.
5. 5Inspect the usage breakdown after the run before using its cost as a team estimate.

A larger window is most useful when the work is connected. If the second request has a different goal, start another session. That keeps each review smaller and avoids paying to carry context that no longer helps. The Cursor model guide gives the same practical rule across models: compare accepted work and review time, not context size by itself.

Choose Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. as the current speed-tier starting point. It succeeds Gemini 3.6 Flash in that role and cuts the published rates by roughly half: input moves from $1.50 to $0.75, cache read from $0.15 to $0.075, and output from $7.50 to $3.50. Both pages list 200k standard context with a 1M maximum.

Choose Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. when its Pro positioning fits the work, especially UI implementation from visual references and visual code analysis. Cursor rates 3.1 Pro medium for speed and cost and frontier for intelligence. Its standard rates are $2 input, $0.20 cache read, and $12 output per million tokens. That is a different trade rather than an automatic upgrade for every task.

Run the same small task in fresh sessions when the choice matters. Keep the prompt, files, and test fixed, then compare the final diff, failed attempts, elapsed time, and usage. The 3.7 rate advantage is clear from the table. Whether its result is better on your repository is something the rate card cannot answer.

Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. can use Cursor's full agent tool set. It can search files and folders, fetch web results, read and edit files, run shell commands, control the browser, generate images, ask clarifying questions, and fetch rules. The model page lists these as available capabilities, while the current product surface, sandbox, approvals, credentials, and team settings still decide what a particular run can do.

Tool access should change the prompt you write. Ask for the verification step alongside the implementation: run the named test, inspect the target route, or show the final file. A model with browser and shell access can collect better evidence than a chat response, but only when the task asks for it. External writes and production changes still need the authority that applies to that system.

Cursor Router can route Auto requests to models in its changing pool, but Cursor does not promise that a particular request will use Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition.. Router is available on Teams and Enterprise. Cost keeps bundled Auto pricing, while Balance and Intelligence charge the rate of the model actually used. Third-party routes also receive the Cursor Token Rate.

Select Gemini 3.7 FlashGoogle's latest speed-tier model in Cursor for high-throughput coding, with a 1M maximum context window and a 90% cached-input discount. Press Enter for the full definition. directly when the model identity is part of the test or approval. Use Router when the goal is to let Cursor choose per request and the team accepts that the underlying pool changes. If admins expose the routed model in responses, review that record before attributing a result or cost to Gemini 3.7 Flash.