# Gemini 3 Flash

> Gemini 3 reasoning at Flash speed and cost — and it beats Gemini 3 Pro on agentic coding.

- Maker: Google
- Line: Gemini Flash
- Canonical page: https://ai-tldr.dev/models/gemini-3-flash/
- Updated: 2026-06-18

## In short

Gemini 3 Flash is Google's fast, cost-efficient model in the Gemini 3 family, released on December 17, 2025. It has a 1M-token context window and costs $0.50 per million input tokens and $3.00 per million output tokens. You can use it through Google AI Studio / Gemini API, Google Vertex AI and OpenRouter.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 1M (max output 64K) | [1] |
| Input price | $0.50 / 1M tokens | [2] |
| Output price | $3.00 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | 17 December 2025 | [3] |
| Parameters | Not disclosed | [1] |
| Knowledge cutoff | January 2025 | [1] |

## Sources

1. [Gemini 3 Flash — Vertex AI model card](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-flash)
2. [Introducing Gemini 3 Flash — Google blog](https://blog.google/products/gemini/gemini-3-flash/)
3. [Gemini 3 Flash on Google DeepMind](https://deepmind.google/models/gemini/)

## Overview

Gemini 3 Flash is Google's fast, cost-efficient model in the Gemini 3 family, released on December 17, 2025. It carries over the complex reasoning, multimodal understanding, and tool-use abilities of Gemini 3 Pro, but is tuned for low latency and low price — Google positions it as the everyday workhorse for high-volume, latency-sensitive workloads where Pro would be overkill.

Despite being the smaller, cheaper tier, Gemini 3 Flash punches well above its weight. It accepts text, images, audio, video, and PDFs across a 1 million-token context window and returns up to roughly 64K tokens of text output. It supports configurable 'thinking levels' (minimal, low, medium, high), structured output, function calling, and automatic context caching, so developers can dial reasoning effort up or down per request to trade cost against quality.

On Google's published benchmarks, Gemini 3 Flash scores 90.4% on GPQA Diamond and 81.2% on MMMU-Pro — close to Gemini 3 Pro — and reaches 78% on SWE-bench Verified for agentic coding, where Google reports it actually edges out Gemini 3 Pro. At $0.50 per 1M input tokens and $3 per 1M output tokens, it undercuts the Pro tier substantially while running about 3x faster than Gemini 2.5 Pro.

## Benchmarks

- GPQA Diamond: 90.4% ([source](https://blog.google/products/gemini/gemini-3-flash/))
- MMMU-Pro: 81.2% ([source](https://blog.google/products/gemini/gemini-3-flash/))
- SWE-bench Verified: 78% ([source](https://blog.google/products/gemini/gemini-3-flash/))
- Humanity's Last Exam (no tools): 33.7% ([source](https://blog.google/products/gemini/gemini-3-flash/))
- MCP Atlas: 57.4% ([source](https://9to5google.com/2025/12/17/gemini-3-flash-launch/))
- Toolathlon: 49.4% ([source](https://9to5google.com/2025/12/17/gemini-3-flash-launch/))

## Pricing

- Input: $0.50 / 1M tokens per 1M tokens
- Output: $3.00 / 1M tokens per 1M tokens

Audio input is priced at $1.00 / 1M tokens. Automatic context caching can reduce effective cost for repeated content. Pricing is for the Gemini API; preview model.

Source: https://blog.google/products/gemini/gemini-3-flash/

## Strengths

- Near-Pro reasoning quality at a fraction of the price and latency — about 3x faster than Gemini 2.5 Pro
- Strong agentic coding: 78% on SWE-bench Verified, ahead of Gemini 3 Pro on Google's own figures
- 1 million-token context window for long documents, codebases, and multi-file inputs
- Genuinely multimodal input — text, images, audio, video, and PDFs in a single request
- Configurable thinking levels let you tune cost vs. depth per call, with automatic context caching to cut repeated-prompt costs
- Uses roughly 30% fewer tokens than Gemini 2.5 Pro on typical tasks, lowering effective cost further

## Best for

- High-volume production features where Gemini 3 Pro would be too slow or expensive
- Agentic coding assistants and code-fixing agents (Gemini CLI, Antigravity, IDE tools)
- Long-document and PDF analysis, summarization, and extraction over the 1M-token window
- Multimodal apps that mix images, audio, video, and text in one pipeline
- Real-time chat and search experiences (it powers AI Mode in Search and the Gemini app) needing fast responses
- Tool-using and multi-step workflow agents that need solid reasoning at low latency

## How to access

- [Google AI Studio / Gemini API](https://ai.google.dev/gemini-api/docs/models) — model id `gemini-3-flash-preview`
- [Google Vertex AI](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-flash) — model id `gemini-3-flash`
- [OpenRouter](https://openrouter.ai/google/gemini-3-flash-preview) — model id `google/gemini-3-flash-preview`

## Gemini Flash versions

- [Gemini 3.5 Flash](https://ai-tldr.dev/models/gemini-3-5-flash/) (19 May 2026)
- Gemini 3 Flash (17 December 2025) — this page
- [Gemini 2.5 Flash](https://ai-tldr.dev/models/gemini-2-5-flash/) (17 April 2025)
- [Gemini 2.0 Flash](https://ai-tldr.dev/models/gemini-2-0-flash/) (30 January 2025)
- [Gemini 1.5 Flash](https://ai-tldr.dev/models/gemini-1-5-flash/) (14 May 2024)

## FAQ

### When was Gemini 3 Flash released?

Google released Gemini 3 Flash on December 17, 2025, rolling it out to the Gemini app, AI Mode in Search, and via API through Google AI Studio, Vertex AI, Google Antigravity, and Gemini Enterprise. It launched as a preview model.

### How much does Gemini 3 Flash cost?

Via the Gemini API it costs $0.50 per 1M input tokens and $3.00 per 1M output tokens, with audio input priced at $1.00 per 1M tokens. Automatic context caching can lower the effective cost for repeated content.

### How does Gemini 3 Flash compare to Gemini 3 Pro?

Gemini 3 Flash is the faster, cheaper tier — roughly 3x faster than Gemini 2.5 Pro — yet stays close to Gemini 3 Pro on reasoning (90.4% GPQA Diamond, 81.2% MMMU-Pro). On Google's figures it actually beats Gemini 3 Pro on agentic coding, scoring 78% on SWE-bench Verified.

### What inputs does Gemini 3 Flash support?

It accepts text, images, audio, video, and PDFs across a 1 million-token context window, and returns text output (up to about 64K tokens). It also supports configurable thinking levels, structured output, and function calling.

## Links

- [Introducing Gemini 3 Flash — Google blog](https://blog.google/products/gemini/gemini-3-flash/)
- [Gemini 3 Flash on Google DeepMind](https://deepmind.google/models/gemini/)
- [Gemini 3 Flash — Vertex AI model card](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-flash)
- [Gemini API pricing](https://ai.google.dev/gemini-api/docs/pricing)
- [Gemini 3 developer guide](https://ai.google.dev/gemini-api/docs/gemini-3)
- [Gemini 3 Flash Preview — OpenRouter](https://openrouter.ai/google/gemini-3-flash-preview)
- [Google announces Gemini 3 Flash — 9to5Google](https://9to5google.com/2025/12/17/gemini-3-flash-launch/)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/gemini-3-flash/