模型坐标

MODEL ANALYSIS

Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning)

Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning)（Google）公开数据详情：智能指数 10.4（v4.3）、上下文 1,000,000 Token。评测来源 Artificial Analysis；规格来源 Artificial Analysis，快照 2026-09-19；各字段保留原始来源与观测时间。

数据快照：2026-09-19查看原始来源 ↗
智能指数 · v4.310.4*暂无排名
生成速度—Token/s
API 价格—当前 API 未核实 · Artificial Analysis model aggregate 历史输入 $0.1 / 输出 $0.4（2026-07-21）
单任务成本 · 历史参考—来源未提供可比较价格
单任务输出 · 历史参考—Token

评测按字段保留来源版本与观测时间。Intelligence Index v4.3 观测于 2026-09-09T14:00:04.717Z。 单任务成本、Token 与完整评测成本仍为此前快照参考，不代表新版评测或当前 API 报价。详情可在指标历史核对。

模型概览

Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning) 的 Intelligence Index v4.3 为 10.4（估算值），生成速度为 暂无数据 Token/s，历史完整评测成本为 暂无数据。这些指标分别反映能力、吞吐和评测经济性；旧成本未与新版指数绑定，不能合并成单一“最好”结论。

技术规格

机构
Google
发布日期
2025-09-08
推理类型
推理模型
模型权重
未开放权重
上下文窗口
1,000,000 Token
输入模态
文本、图像、语音、视频
输出模态
文本
性能来源
Google

INTERACTIVE COMPARISON

模型指标比较器

切换指标并按状态、推理类型和权重开放性筛选。首屏服务端呈现前 24 个模型，浏览器随后载入完整数据；历史参考配置可在“历史”或“全部”中查看。

23 / 653 个模型 · 正在载入全量数据

仅比较 v4.3 同版本观测；不跨方法版本排序。估算指数保留来源的暂定性质。

智能指数模型比较；当前显示 23 个模型。Claude Fable 5.1 (max with …53.4Claude Fable 5.1 (xhigh wit…53.2GPT-6 Astra (max)52.8GPT-6 Astra (xhigh)52.5Claude Fable 5.1 (high with…51.2GPT-6 Astra (high)51Claude Opus 5 (max)50.7Claude Fable 549.7GPT-6 Astra (medium)49.7Claude Opus 5 (xhigh)49.7Claude Fable 5.1 (medium wi…49.1Claude Opus 5 (high)48.2Muse Spark 1.3 (max)48.2GPT-5.6 Sol (max)47.1Claude Fable 5.1 (low with …47GPT-6 Astra (low)46Qwen3.8 Max (0902)45.4Muse Spark 1.3 (xhigh)45.2Claude Opus 5 (medium)45.1GLM-5.344.9Grok 4.6 (high)44.4Grok 4.6 (xhigh)44.3GPT-5.6 Sol (xhigh)44.1

INTELLIGENCE

Artificial Analysis 智能指数 · v4.3

v4.3 按 Agent、编码、通用与科学推理四类能力组合。仅在同版本内比较；数值越高越好。本页配置原观测于 2026-09-09T14:00:04.717Z。蓝色代表当前模型。

模型智能指数对比 · v4.3
模型智能指数对比 · v4.3；公开数据053.4Claude Fable 5.1 (max with fallback): 53.4；AnthropicClaude Fable 5.1 (max with fa…53.4Claude Fable 5.1 (xhigh with fallback): 53.2；AnthropicClaude Fable 5.1 (xhigh with …53.2GPT-6 Astra (max): 52.8；OpenAIGPT-6 Astra (max)52.8GPT-6 Astra (xhigh): 52.5；OpenAIGPT-6 Astra (xhigh)52.5Claude Fable 5.1 (high with fallback): 51.2；AnthropicClaude Fable 5.1 (high with f…51.2GPT-6 Astra (high): 51；OpenAIGPT-6 Astra (high)51Claude Opus 5 (max): 50.7；AnthropicClaude Opus 5 (max)50.7Claude Fable 5: 49.7；AnthropicClaude Fable 549.7GPT-6 Astra (medium): 49.7；OpenAIGPT-6 Astra (medium)49.7Claude Opus 5 (xhigh): 49.7；AnthropicClaude Opus 5 (xhigh)49.7Claude Fable 5.1 (medium with fallback): 49.1；AnthropicClaude Fable 5.1 (medium with…49.1Claude Opus 5 (high): 48.2；AnthropicClaude Opus 5 (high)48.2Muse Spark 1.3 (max): 48.2；MetaMuse Spark 1.3 (max)48.2GPT-5.6 Sol (max): 47.1；OpenAIGPT-5.6 Sol (max)47.1Claude Fable 5.1 (low with fallback): 47；AnthropicClaude Fable 5.1 (low with fa…47GPT-6 Astra (low): 46；OpenAIGPT-6 Astra (low)46Qwen3.8 Max (0902): 45.4；AlibabaQwen3.8 Max (0902)45.4Muse Spark 1.3 (xhigh): 45.2；MetaMuse Spark 1.3 (xhigh)45.2Claude Opus 5 (medium): 45.1；AnthropicClaude Opus 5 (medium)45.1GLM-5.3: 44.9；Z AIGLM-5.344.9
查看图表数据
项目数值
Claude Fable 5.1 (max with fallback)53.4
Claude Fable 5.1 (xhigh with fallback)53.2
GPT-6 Astra (max)52.8
GPT-6 Astra (xhigh)52.5
Claude Fable 5.1 (high with fallback)51.2
GPT-6 Astra (high)51
Claude Opus 5 (max)50.7
Claude Fable 549.7
GPT-6 Astra (medium)49.7
Claude Opus 5 (xhigh)49.7
Claude Fable 5.1 (medium with fallback)49.1
Claude Opus 5 (high)48.2
Muse Spark 1.3 (max)48.2
GPT-5.6 Sol (max)47.1
Claude Fable 5.1 (low with fallback)47
GPT-6 Astra (low)46
Qwen3.8 Max (0902)45.4
Muse Spark 1.3 (xhigh)45.2
Claude Opus 5 (medium)45.1
GLM-5.344.9

INTELLIGENCE BREAKDOWN

组成评测 · v4.3

仅展示该版本逐项核实的原始指标；缺少同版本观测的字段不从历史回填。Elo、指数与通过率不能当作同一量纲排序；幻觉率越低越好。

Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning) 的分项原始指标
指标原始成绩
AA-Omniscience 准确率18.0%
AA-Omniscience 幻觉率（越低越好）87.8%
AA-LCR v1.164.7%
Humanity's Last Exam7.0%
CritPt0.0%

AA-OMNISCIENCE

知识与幻觉

指数同时考虑正确回答和避免幻觉。当前模型准确率 18.0%，幻觉率 87.8%。

AA-Omniscience 指数对比
AA-Omniscience 指数对比；公开数据043.7GPT-6 Astra (high): 43.7；OpenAIGPT-6 Astra (high)43.7Claude Fable 5.1 (max with fallback): 43.5；AnthropicClaude Fable 5.1 (max with fa…43.5GPT-6 Astra (xhigh): 43.4；OpenAIGPT-6 Astra (xhigh)43.4GPT-6 Astra (max): 43.4；OpenAIGPT-6 Astra (max)43.4Claude Fable 5: 43.3；AnthropicClaude Fable 543.3Claude Fable 5.1 (xhigh with fallback): 42.4；AnthropicClaude Fable 5.1 (xhigh with …42.4GPT-6 Astra (medium): 42.2；OpenAIGPT-6 Astra (medium)42.2Claude Fable 5.1 (high with fallback): 40.8；AnthropicClaude Fable 5.1 (high with f…40.8GPT-6 Astra (low): 40.6；OpenAIGPT-6 Astra (low)40.6Claude Fable 5.1 (medium with fallback): 37.6；AnthropicClaude Fable 5.1 (medium with…37.6Claude Opus 5 (max): 37.1；AnthropicClaude Opus 5 (max)37.1Claude Opus 5 (xhigh): 35.4；AnthropicClaude Opus 5 (xhigh)35.4Claude Fable 5.1 (low with fallback): 34.1；AnthropicClaude Fable 5.1 (low with fa…34.1Claude Opus 5 (high): 33.7；AnthropicClaude Opus 5 (high)33.7Gemini 3.1 Pro Preview: 31.9；GoogleGemini 3.1 Pro Preview31.9Claude Opus 5 (medium): 31；AnthropicClaude Opus 5 (medium)31Grok 4.6 (high): 30.5；SpaceXAIGrok 4.6 (high)30.5Gemini 3.8 Flash (high): 29.6；GoogleGemini 3.8 Flash (high)29.6Grok 4.6 (xhigh): 29.3；SpaceXAIGrok 4.6 (xhigh)29.3Gemini 3.8 Flash (medium): 28.6；GoogleGemini 3.8 Flash (medium)28.6
查看图表数据
项目数值
GPT-6 Astra (high)43.7
Claude Fable 5.1 (max with fallback)43.5
GPT-6 Astra (xhigh)43.4
GPT-6 Astra (max)43.4
Claude Fable 543.3
Claude Fable 5.1 (xhigh with fallback)42.4
GPT-6 Astra (medium)42.2
Claude Fable 5.1 (high with fallback)40.8
GPT-6 Astra (low)40.6
Claude Fable 5.1 (medium with fallback)37.6
Claude Opus 5 (max)37.1
Claude Opus 5 (xhigh)35.4
Claude Fable 5.1 (low with fallback)34.1
Claude Opus 5 (high)33.7
Gemini 3.1 Pro Preview31.9
Claude Opus 5 (medium)31
Grok 4.6 (high)30.5
Gemini 3.8 Flash (high)29.6
Grok 4.6 (xhigh)29.3
Gemini 3.8 Flash (medium)28.6

CONTEXT WINDOW

上下文窗口

上下文窗口表示单次请求可处理的最大 Token 数，不代表模型一定能在全窗口内保持同等检索质量。

上下文窗口对比
上下文窗口对比；公开数据010,000,000 TokenLlama 4 Scout: 10,000,000 Token；MetaLlama 4 Scout10,000,000 TokenMuse Spark 1.2 (xhigh): 1,048,576 Token；MetaMuse Spark 1.2 (xhigh)1,048,576 TokenKimi K3 (low): 1,048,576 Token；KimiKimi K3 (low)1,048,576 TokenSolar Open2 250B: 1,048,576 Token；UpstageSolar Open2 250B1,048,576 TokenKimi K3 (max): 1,048,576 Token；KimiKimi K3 (max)1,048,576 TokenMiMo-V2.5-Pro: 1,000,000 Token；XiaomiMiMo-V2.5-Pro1,000,000 TokenNova 2.0 Omni (medium): 1,000,000 Token；AmazonNova 2.0 Omni (medium)1,000,000 TokenGemini 3.5 Flash (minimal): 1,000,000 Token；GoogleGemini 3.5 Flash (minimal)1,000,000 TokenGPT-5.6 Luna (low): 1,000,000 Token；OpenAIGPT-5.6 Luna (low)1,000,000 TokenGPT-5.6 Sol (low): 1,000,000 Token；OpenAIGPT-5.6 Sol (low)1,000,000 TokenGemini 3.5 Flash-Lite: 1,000,000 Token；GoogleGemini 3.5 Flash-Lite1,000,000 TokenInkling: 1,000,000 Token；Thinking MachinesInkling1,000,000 TokenClaude Opus 5 (xhigh): 1,000,000 Token；AnthropicClaude Opus 5 (xhigh)1,000,000 TokenClaude Sonnet 5 (low): 1,000,000 Token；AnthropicClaude Sonnet 5 (low)1,000,000 TokenGLM-5.3-Flash: 1,000,000 Token；Z AIGLM-5.3-Flash1,000,000 TokenNova 2.0 Omni: 1,000,000 Token；AmazonNova 2.0 Omni1,000,000 TokenClaude Opus 5 (low): 1,000,000 Token；AnthropicClaude Opus 5 (low)1,000,000 TokenMiMo-V2.5: 1,000,000 Token；XiaomiMiMo-V2.51,000,000 TokenNemotron 3 Nano: 1,000,000 Token；NVIDIANemotron 3 Nano1,000,000 TokenClaude Sonnet 5 (max): 1,000,000 Token；AnthropicClaude Sonnet 5 (max)1,000,000 Token
查看图表数据
项目数值
Llama 4 Scout10,000,000 Token
Muse Spark 1.2 (xhigh)1,048,576 Token
Kimi K3 (low)1,048,576 Token
Solar Open2 250B1,048,576 Token
Kimi K3 (max)1,048,576 Token
MiMo-V2.5-Pro1,000,000 Token
Nova 2.0 Omni (medium)1,000,000 Token
Gemini 3.5 Flash (minimal)1,000,000 Token
GPT-5.6 Luna (low)1,000,000 Token
GPT-5.6 Sol (low)1,000,000 Token
Gemini 3.5 Flash-Lite1,000,000 Token
Inkling1,000,000 Token
Claude Opus 5 (xhigh)1,000,000 Token
Claude Sonnet 5 (low)1,000,000 Token
GLM-5.3-Flash1,000,000 Token
Nova 2.0 Omni1,000,000 Token
Claude Opus 5 (low)1,000,000 Token
MiMo-V2.51,000,000 Token
Nemotron 3 Nano1,000,000 Token
Claude Sonnet 5 (max)1,000,000 Token

FAQ

Gemini 2.5 Flash-Lite 常见问题

问题顺序参考公开搜索联想；回答只使用本页可核对的来源数据。

Gemini 2.5 Flash-Lite 什么时候发布？

Artificial Analysis 当前记录的发布日期为 2025-09-08；本站按来源快照展示，不把发布前预测当作事实。

Gemini 2.5 Flash-Lite API 价格是多少？

当前没有精确同款的已核实 API 报价；本站上次记录的 Artificial Analysis model aggregate 数据为输入 US$0.1、输出 US$0.4（2026-07-21），仅作历史参考。

Gemini 2.5 Flash-Lite 的评测表现怎么样？

公开 Intelligence Index v4.3 为 10.4，该数值为估算值。观测于 2026-09-09T14:00:04.717Z。不同方法版本不直接比较，不等于所有任务中的统一胜率。

Gemini 2.5 Flash-Lite 的上下文窗口有多大？

上下文窗口为 1,000,000 Token；窗口上限不代表模型在全长度上始终保持相同的检索与推理质量。

Gemini 2.5 Flash-Lite 的生成速度是多少？

当前来源未提供生成速度。

Gemini 2.5 Flash-Lite 是开源模型吗？

当前Artificial Analysis未提供可核验的开放权重，本站暂按非开放权重处理。

怎样通过 API 使用 Gemini 2.5 Flash-Lite？

原始来源记录了 1 家 Provider。 Provider 横向明细尚未同步。

RELATED PATHS

继续比较模型与任务