Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.
Additionally, Gemini 2.5 Flash is configurable through the "max tokens for reasoning" parameter, as described in the documentation (https://dynamicsinfo.com/docs/use-cases/reasoning-tokens#max-tokens-for-reasoningOpens in new tab).
提供商
不同公司托管同一模型。OpenRouter 根据您选择的路由模式将请求路由到其中一家公司:Balanced(价格 + 速度)、Nitro(速度最快)或 Floor(价格最低)。
| US$0.30 | US$2.50 | US$0.030 | US$0.10 | 0.69s | 107 tok/s | ||
不用于 Standard 路由:未使用这些端点的原因 | |||||||
Flex | US$0.15 | US$1.25 | US$0.015 | US$0.05 | 0.97s | 206 tok/s | |
Priority | US$0.54 | US$4.50 | US$0.054 | US$0.18 | 0.83s | 107 tok/s | |
- 输入 US$0.30 / 输出 US$2.50107 tok/s99.96% 正常运行时间
- 不用于 Standard 路由:未使用这些端点的原因
- 输入 US$0.15 / 输出 US$1.25206 tok/s99.92% 正常运行时间Flex
- 输入 US$0.54 / 输出 US$4.50107 tok/s93.66% 正常运行时间Priority
定价
客户为此模型实际支付的平均价格,与提供商公布的价格并列展示。由于缓存和折扣,实际支付的价格往往远低于标价。
加权平均值
Gemini 2.5 Flash — 价格历史
| US$0.1940 | US$2.50 | 40.9% | 99.3% | ||
不用于 Standard 路由:未使用这些端点的原因 | |||||
Flex | US$0.0698 | US$1.25 | 59.4% | 0.5% | |
Priority | US$0.2769 | US$4.49 | 54.2% | 0.2% | |
- 有效输入价格 US$0.1940 / 实际输出 US$2.50token 占比99.3%
- 不用于 Standard 路由:未使用这些端点的原因
- 有效输入价格 US$0.0698 / 实际输出 US$1.25Flextoken 占比0.5%
- 有效输入价格 US$0.2769 / 实际输出 US$4.49Prioritytoken 占比0.2%
性能
吞吐量是模型生成内容的速度(每秒 token 数,越高越好)。延迟是完整的往返时间(越低越好)。TTFT 是首个 token 的响应时间,即您看到内容出现前需要等待的时间(越低越好)。
P50,所有提供商中最佳
P50,最佳提供商
正常运行时间
正常运行时间是指过去 3 天中至少有一家提供商响应请求的时间百分比。可用性是指成功提供推理服务的时间百分比。OpenRouter 会持续监控,并在某家提供商返回错误时使用次优提供商。
过去 3 天的可用性
最近 72 小时过去 24 小时的可用性
当上游提供商发生错误时,如果您的请求筛选条件允许,我们可以通过路由到另一家健康的提供商来恢复服务。您可以通过以下接口以编程方式访问各提供商的正常运行时间数据: 端点 API。了解更多有关我们的 负载均衡和提供商路由选项。
基准测试
标准化评测得分,以及此模型针对共用基准测试提示词的实际输出。百分比越高越好,每项排名显示此模型在该来源所评测的模型中的位置。
| 来源 | 基准测试 | 得分 |
|---|---|---|
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) GPQA Diamond | 68.3% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) HLE | 4.7% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) IFBench | 39.0% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) τ²-Bench Telecom | 14.9% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) AA-LCR | 49.9% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) CritPt | 1.4% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) LiveCodeBench | 49.5% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) Terminal-Bench Hard | 12.1% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) MMLU-Pro | 80.9% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) AIME 2025 | 60.3% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) AIME | 50.0% |
| Artificial Analysis | Gemini 2.5 Flash (Non-reasoning) MATH-500 | 93.2% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) GPQA Diamond | 79.0% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) HLE | 12.1% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) IFBench | 50.3% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) τ²-Bench Telecom | 31.6% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) AA-LCR | 65.3% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) CritPt | 1.1% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) LiveCodeBench | 69.5% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) Terminal-Bench Hard | 13.6% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) MMLU-Pro | 83.2% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) AIME 2025 | 73.3% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) AIME | 82.3% |
| Artificial Analysis | Gemini 2.5 Flash (Reasoning) MATH-500 | 98.1% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) GPQA Diamond | 76.6% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) HLE | 8.7% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) IFBench | 43.5% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) τ²-Bench Telecom | 28.4% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) AA-LCR | 60.0% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) CritPt | 0.0% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) LiveCodeBench | 62.5% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) Terminal-Bench Hard | 14.4% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) MMLU-Pro | 83.6% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) AIME 2025 | 56.7% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) GPQA Diamond | 79.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) HLE | 13.8% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) IFBench | 52.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) τ²-Bench Telecom | 45.6% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) AA-LCR | 71.0% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) CritPt | 0.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) LiveCodeBench | 71.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) Terminal-Bench Hard | 16.7% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) MMLU-Pro | 84.2% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Sep '25) (Reasoning) AIME 2025 | 78.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) GPQA Diamond | 59.4% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) HLE | 3.8% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) LiveCodeBench | 40.6% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) MMLU-Pro | 78.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) AIME | 43.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Non-reasoning) MATH-500 | 92.6% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) GPQA Diamond | 69.8% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) HLE | 12.1% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) LiveCodeBench | 50.5% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) MMLU-Pro | 80.0% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) AIME | 84.3% |
| Artificial Analysis | Gemini 2.5 Flash Preview (Reasoning) MATH-500 | 98.1% |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena 3D Elo | 1051 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena Code Categories Elo | 1093 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena Data Visualization Elo | 1130 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena Game Development Elo | 1055 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena SVG Elo | 1008 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena UI Component Elo | 1083 |
| Design Arena | Gemini 2.5 Flash Preview 09-2025 Models Arena Website Elo | 1107 |
| Vals AI | Gemini 2.5 Flash (7/17) (Nonthinking) IOI v1 | 2.6% |
| Vals AI | Gemini 2.5 Flash (7/17) (Nonthinking) MedCode | 38.4% |
| Vals AI | Gemini 2.5 Flash (7/17) (Nonthinking) MedScribe | 82.9% |
| Vals AI | Gemini 2.5 Flash (7/17) (Nonthinking) SAGE v1 | 44.8% |
应用
向此模型发送最多流量的公开应用。这能很好地反映实际生产工作负载的情况,也能提示此模型最适合哪些用例。
活动
此模型的 token 用量和请求流量随时间的变化。
API
可直接用于调用此模型的代码。OpenRouter 的 API 与 OpenAI 兼容——大多数 SDK 只需替换基础 URL 即可使用。切换模型时,唯一需要更改的是下方的模型标识符。
目前无法通过 EU 区域内路由使用。 区域内路由文档
文档
在 OpenRouter 上使用 Google: Gemini 2.5 Flash 构建应用的指南、实用教程、博客文章和工具。
常见问题
Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.
OpenRouter 上的 Gemini 2.5 Flash 文档包括 Gemini 2.5 Flash API - Pricing, Quickstart & Provider Comparison 博客文章。快速入门介绍了 API 密钥和您的首次请求,API 参考列出了所有参数。
Gemini 2.5 Flash 的费用为 US$0.30/百万输入 token 和 US$2.50/百万输出 token,Cache Read,价格为 US$0.03/M tokens、Cache Write,价格为 US$0.08333/M tokens、Image Input,价格为 US$0.30/M tokens、Input Audio,价格为 US$1.00/M tokens和Input Audio Cache,价格为 US$0.10/M tokens 单独计费。
Gemini 2.5 Flash 的上下文窗口为 1,048,576 个 token。它最多支持 65,535 个补全 token。
是的。Gemini 2.5 Flash 接受 tools 和 tool_choice,以进行函数调用。 它还支持通过 response_format 中的 JSON schema 实现结构化输出。
Gemini 2.5 Flash 接受 PDF 等文件、图像、文本、音频和视频 作为输入,并返回 文本。
在 OpenRouter 上,Gemini 2.5 Flash 由 1 家提供商提供服务:Google AI Studio。
Gemini 4 Hydrogen, Gemini Decision 1, Gemini 4 Carbon 及另外 31 个 是 Google 的其他 文本 模型。
Gemini 2.5 Flash 于 2025年6月17日 发布。其知识截止日期为 2025年1月31日。