Gemini 3.6 Flash
Gemini 3.6 Flash is a lightweight-tier model in Google's Gemini family, accepting text, image, video, and audio input across a 1,048,576-token context window.
API Pricing
Cheapest input on Google Cloud · cheapest output on Google Cloud| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $0.375 | $1.88 | $0.037 | |
| $0.375 | $1.88 | $0.037 | |
| $0.750 | $3.75 | $0.075 | |
| $0.750 | $3.75 | $0.075 | |
| $1.50 | $7.50 | $0.150 | |
| $1.50 | $7.50 | - |
Prices updated daily. Last check: Oct 26, 2026
Compare API pricing for every Gemini model →Gemini 3.6 Flash pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- GPQA Diamond92.8%
- Humanity's Last Exam40.8%
Coding
- SciCode53.4%
Agentic & Tool Use
- Terminal-Bench v2.177.5%
- τ-bench Banking29.9%
Instruction & Long Context
- Long-Context Reasoning80.0%
Benchmarks measured Oct 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- Family
- Gemini
- Tier
- Lightweight
- Context Window
- 1.0M
- Modalities
- Text, Image, Video, Audio
Capabilities
- Tool Calling
- Yes
- Open Source
- No
- Aliases
- gemini-3.6-flash, Gemini 3.6 Flash, models/gemini-3.6-flash
Strengths & Limitations
Strengths
- 1,048,576-token context window, enough for very large document sets, codebases, or long transcripts in a single request
- Accepts four input modalities: text, image, video, and audio
- Supports tool calling, enabling function-calling and agentic workflows
- Measured output throughput of roughly 209 tokens per second (Artificial Analysis), suited to streaming-heavy use
- Lightweight Flash tier within the Gemini family, positioned for high-volume workloads rather than one-off queries
- Native video and audio input avoids separate transcription or frame-extraction preprocessing steps
Limitations
- Time to first token measured at about 13.6 seconds, which is poor for interactive chat or autocomplete-style latency requirements
- Lightweight tier — Pro-tier Gemini models are generally the choice for the hardest reasoning and coding tasks
- We do not track published task benchmark scores (e.g. coding or reasoning suites) for this model, so capability comparisons are limited
- Filling a one-million-token context consumes a large number of input tokens per call, which affects cost regardless of the per-token rate
- Weight availability and licensing are not something we track for this model
Key Features
About Gemini 3.6 Flash
Common Use Cases
Gemini 3.6 Flash fits workloads that combine large or multimodal inputs with high request volume: summarizing and querying long documents, contracts, or logs; analyzing video and audio recordings such as meetings, lectures, or support calls without a separate transcription stage; extracting structured fields from mixed text-and-image sources; and batch classification, tagging, or moderation over large corpora. Tool calling makes it usable as the execution model inside agent loops and retrieval pipelines where many short tool-mediated steps are issued. Its measured time to first token is high, so it is a weaker fit for interactive chat, IDE completions, or anything with a tight first-response budget; for those, a model with lower startup latency is preferable, while Pro-tier Gemini models remain the more common pick for the most demanding reasoning and coding tasks.
Frequently Asked Questions
How much does Gemini 3.6 Flash cost?
Pricing depends on the provider and the pricing model — input versus output tokens, cached input, batch versus real-time, and any modality-specific rates. Because rates change frequently and differ between providers, check the pricing table on this page for current figures rather than relying on a fixed number.
What is Gemini 3.6 Flash best used for?
It suits high-volume and long-input work: long-document and transcript analysis, video and audio understanding, structured extraction from mixed text and image inputs, batch classification, and tool-calling agent steps. Its million-token context and four-modality input surface are the main reasons to choose it over a smaller text-only model.
How does it compare to Pro-tier Gemini models?
Gemini 3.6 Flash sits at the lightweight tier of the Gemini family, which is generally oriented toward throughput and volume, while Pro-tier models are the ones typically selected for the hardest reasoning, coding, and multi-step analytical tasks. If your workload is many similar requests over large inputs, Flash is the tier designed for it.
Is it a good fit for real-time chat?
Its measured time to first token is around 13.6 seconds (Artificial Analysis), which is high for interactive use even though output then streams at roughly 209 tokens per second. For latency-sensitive chat or code completion, a model with faster first-token behavior is a better match; for batch or background jobs the startup delay matters much less.
Can it process video and audio directly?
Yes — the model's confirmed input modalities include text, image, video, and audio, so media can be passed in without a separate transcription or frame-extraction step. Practical per-request limits on media length and size are set by the serving provider.