The Gemini built for volume — a million-token window, thinking when you want it, and pricing that survives contact with production traffic.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Gemini 2.5 Flash is the model you reach for when the job runs a million times, not once.
It keeps the thing that made Gemini 2.5 Pro interesting — a 1,048,576-token context window — and gives up deliberation rather than capacity. That is an unusual trade. Most fast models shrink the window; Flash keeps it and shortens the thinking instead.
Flash's knowledge cutoff is January 2025. For a model serving production traffic today, that is a long time ago, and it is the single fact most teams discover too late.
Google's answer is grounding with Google Search, which Flash supports. Turn it on and the model retrieves current information rather than reciting a stale parameter memory. Leave it off and ask about anything recent, and you will get confident, dated answers. Decide this deliberately — it is not a default you want to inherit by accident.
Across providers Flash lands near $0.30 per million input tokens and $2.50 per million output tokens, though this moves by provider and region and should be confirmed against your own account before you model a budget.
Two features matter more than the headline rate:
Thinking is the lever in the other direction. Flash supports it, and a generous thinking budget erodes exactly the price advantage you chose Flash for.
Input is genuinely multimodal: text, images, audio, video and PDFs, with no extraction step in front of the model. A support pipeline can pass a screenshot and a call recording together and ask one question about both.
Output is text. Full stop. Flash does not generate images, does not synthesise speech, and does not support the Live API for real-time bidirectional sessions. If either half of your product needs generated media, that half needs a different model.
Code execution and structured outputs are both supported, so it slots into agent loops and typed pipelines without prompt-engineering the response shape by hand.
Flash answers. Pro reasons.
For extraction, routing, classification, tagging and summarisation — tasks with a findable correct answer — Flash is the right call and the latency difference is free money. For work where a wrong answer is expensive and the path to the right one has several steps, Gemini 2.5 Pro earns its slower response.
| Field | Value |
|---|---|
| Provider | |
| Context window | 1,048,576 tokens |
| Maximum output | 65,536 tokens |
| Knowledge cutoff | January 2025 |
| Input | text, image, audio, video, PDF |
| Output | text |
| Thinking | Supported |
| Function calling | Supported |
| Structured output | Supported |
| Context caching | Supported |
| Code execution | Supported |
| Search grounding | Supported |
| Batch API | Supported |
| Image / audio generation | Not supported |
| Released | June 2025 |
Pricing and regional availability move often. Treat the figures above as indicative and confirm them against your own account before you commit to a budget.
Compare Flash against everything else in the AI model directory.
Use the recorded facts as a shortlist, then validate the model against representative inputs and production constraints.
Gemini 2.5 Flash is categorized for Audio-text-to-text, Automatic speech recognition, Image-text-to-text, Image-to-text, Text generation, Text-to-text AI models, Video-text-to-text. Its current record accepts text, image, audio, video, pdf and produces text. Confirm file formats, preprocessing, and provider-specific request schemas before implementation.
The directory records 1.0M context and 65.5K maximum output. Listed token prices are $0.30 input and $2.50 output per 1M tokens. Treat missing values as unknown and recheck current commercial terms.
Recorded output speed is — and time to first token is —. Hosting route, region, prompt length, concurrency, and provider load can materially change both measurements.
This page separates sourced model facts from incomplete fields. Benchmark evidence is displayed only when its source is linked and verified. Before choosing Gemini 2.5 Flash, test task quality, tool reliability, safety behavior, data controls, rate limits, and total cost on the exact route you intend to use.
Related approved models in Automatic speech recognition. Compare specifications and verify fit for your workload.
Google's low-cost Gemini 3 model with a 1,048,576-token context window, vision support, and $0.30/$2.50 per million token pricing for volume workloads.
Google's mid-tier Gemini 3 model: 1,048,576-token context, multimodal input, and $1.50/$7.50 per million token pricing for production agent workloads.
Google's cheapest Gemini 3 model, with a 1,048,576-token context window and $0.25/$1.50 per million token pricing for high-volume, cost-sensitive work.
Google's full Gemini 3 Flash model, with a 1,048,576-token context window and $1.50/$9.00 per million token pricing for production-grade agent work.
Google's fast, rolling-updated Gemini model with a 1M-token context window, reasoning, and tool calling at $1.50/$9.00 per million tokens.
Google's low-cost, high-volume Gemini model at $0.25/$1.50 per million tokens, with a 1M-token context window and reasoning support.