Loading page
Google
Gemini 3.5 Flash
Gemini 3.5 Flash is a model in Google's Gemini Flash line, the family's speed-and-efficiency tier, measured at roughly 260 output tokens per second in third-party benchmarking.
Input from
$0.600 / 1M tokens
across 5 providers
API Pricing
Cheapest on Velokey — 48% below avg| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $0.600 | $3.60 | $0.060 | |
| $0.750 | $4.50 | $0.075 | |
| $0.750 | $4.50 | $0.075 | |
| $1.50 | $9.00 | - | |
| $1.50 | $9.00 | $0.150 | |
| $1.50 | $9.00 | $0.150 | |
| $1.50 | $9.00 | $0.150 |
Prices updated daily. Last check: Sep 21, 2026
Gemini 3.5 Flash pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Intelligence
33.6 / 100
Reasoning & Knowledge
- GPQA Diamond92.1%
- Humanity's Last Exam41.3%
Agentic & Tool Use
- Terminal-Bench Hard39.4%
- τ²-bench95.6%
Instruction & Long Context
- IFBench74.6%
- Long-Context Reasoning74.3%
Benchmarks measured Sep 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- Modalities
- Text
Capabilities
- Tool Calling
- No
- Open Source
- No
Strengths & Limitations
Strengths
- Measured at approximately 260 output tokens per second in Artificial Analysis benchmarking, a high generation rate for streaming and long-form output
- Sits in Google's Flash line, the tier the Gemini family uses for throughput- and efficiency-oriented workloads
- Fast decode speed reduces total completion time on long responses relative to slower-generating models
- Part of the Gemini series, so it can typically be swapped in for other Gemini models with minimal client-side change
- Backed by Google's serving infrastructure, which is generally provisioned for high-volume API traffic
Limitations
- Time to first token measured at roughly 11.6 seconds, which makes it a poor fit for latency-sensitive interactive turns with short outputs
- We do not currently track a confirmed context window size for this model
- We do not track confirmed input modality support, so multimodal use should be verified in Google's documentation
- Flash-tier models in general are tuned for efficiency rather than maximum reasoning depth, so the heaviest reasoning tasks may warrant a larger sibling model
- Limited independently verified benchmark coverage in our database beyond throughput and latency measurements
Key Features
•Google Gemini family model
•Flash tier — throughput- and efficiency-oriented line
•~260 output tokens per second measured generation rate (Artificial Analysis)
•~11.6 second measured time to first token (Artificial Analysis)
•Available through cloud inference API providers listed in the pricing table
•API-compatible with the broader Gemini model series
About Gemini 3.5 Flash
Gemini 3.5 Flash is a large language model from Google, positioned in the Gemini family's Flash line. Across Gemini generations, Flash models have served as the throughput-oriented tier — intended for workloads where response speed and cost per token matter more than squeezing out the last increment of reasoning depth. Gemini 3.5 Flash continues that naming convention within the Gemini series.
The measurements we track for this model come from Artificial Analysis: an output generation rate of approximately 260 tokens per second and a time to first token of roughly 11.6 seconds. The high generation rate paired with a long time to first token is a pattern typically seen when a model performs internal deliberation before emitting visible output, so end-to-end latency on short prompts may be dominated by that initial delay rather than by decode speed. For long generations, the throughput figure has a larger effect on total completion time.
Beyond creator, family placement, and these throughput measurements, we do not currently track confirmed specifications for Gemini 3.5 Flash — including context window size, supported input modalities, or API feature details. Readers evaluating the model for a specific integration should confirm those details against Google's own documentation. Provider availability and current rates for this model are shown in the pricing table on this page.
Common Use Cases
Gemini 3.5 Flash's measured profile — high output token throughput with a long time to first token — points toward batch and long-output workloads rather than snappy conversational turns. Good candidates include document summarization, bulk content generation, data extraction and transformation across large record sets, and background agent steps where a few seconds of initial delay is acceptable but total generation time matters. As a Flash-tier model it is the kind of option teams typically reach for when running high request volumes and wanting to reserve larger Gemini models for the subset of requests that need deeper reasoning. For chat interfaces where perceived responsiveness on short replies is the priority, the measured first-token latency is a consideration worth testing against your own prompts.
Frequently Asked Questions
How much does Gemini 3.5 Flash cost?
Pricing varies by provider and by pricing type — input versus output tokens, and whether a provider offers batch or cached-input rates. Rates also change frequently. Check the pricing table on this page for the current figures from each provider we track.
What is Gemini 3.5 Flash best used for?
Its measured throughput of roughly 260 output tokens per second suits long-output and high-volume work: summarization, bulk extraction and classification, content generation, and background agent steps. The measured time to first token of about 11.6 seconds makes it less suited to interactive chat where a fast first word matters.
How does Gemini 3.5 Flash differ from other Gemini models?
It belongs to Google's Flash line, which within the Gemini family is the tier oriented toward speed and efficiency rather than maximum capability. Larger Gemini tiers are generally the choice for the hardest reasoning and agentic tasks, while Flash models target volume workloads. Compare the specific models listed on this site for the measurements we have for each.
What context window does Gemini 3.5 Flash support?
We do not currently have a confirmed context window figure for this model in our database. Google's official model documentation is the authoritative source for context limits and any input modality support.
Why is the time to first token so much higher than the generation speed suggests?
A long time to first token alongside a fast decode rate is a pattern commonly associated with models that deliberate internally before producing visible output. The practical effect is that short responses feel slow while long responses complete comparatively quickly. Both figures come from Artificial Analysis measurements and will vary with prompt length, provider, and load.