跳转到主要内容

Gemini 3 Flash Preview (Non-reasoning)

GoogleGeminiProprietary

描述

Gemini 3 Flash offers frontier intelligence built for speed at a fraction of the cost. It combines Gemini 3's Pro-grade reasoning with Flash-level latency, efficiency and cost. Features a 1 million-token input context window and is optimized for agentic workflows, coding, and complex analysis.

发布日期
2025-12-17
参数规模
上下文长度
1.0M
支持模态
audio, file, image, text, video

能力雷达图

48
general
54
coding
60
reasoning
57
science估算
60
agents
80
multimodal

Science 在缺少专门科学评测时使用推理能力代理估算。

排行榜排名

领域#排名分数来源
智能体与工具25
65.0
LS
代码能力榜74
66.0
AA
通用能力榜118
63.0
AA
数学推理160
56.0
AA
多模态榜71
41.0
LS
推理能力97
34.0
LS
科学能力64
67.0
AA

基准测试分数 (LLM Stats)

Agents

Vending-Bench 2363500.0%自报
t2-bench90.2%自报
MCP Atlas57.4%自报
Toolathlon49.4%自报
Terminal-Bench 2.047.6%自报

Biology

GPQA90.4%自报

Code

LiveCodeBench Pro2316.00 / 3000自报
SWE-Bench Verified78.0%自报

Factuality

SimpleQA68.7%自报
FACTS Grounding61.9%自报

General

Global PIQA92.8%自报
MMMLU91.8%自报
MMMU-Pro81.2%自报
MRCR v2 (8-needle)22.1%自报

Grounding

ScreenSpot Pro69.1%自报

Healthcare

VideoMMMU86.9%自报

Math

AIME 202599.7%自报
Humanity's Last Exam43.5%自报

Multimodal

CharXiv-R80.3%自报
OmniDocBench 1.512.1%自报

Reasoning

ARC-AGI v233.6%自报

AA 评测指数

Math Index
55.7
Coding Index
37.8
Intelligence Index
35.0
Mmlu Pro
0.9
Gpqa
0.8
Livecodebench
0.8
Aime 25
0.6
Ifbench
0.6
Scicode
0.5
Lcr
0.5
Tau2
0.4
Terminalbench Hard
0.3
Hle
0.1

LLM Stats 分类评分

Agents
100
Code
100
Reasoning
100
General
100
Biology
90
Chemistry
90
Language
90
Physics
90
Frontend Development
80
Math
80
Factuality
70
Grounding
70
Multimodal
70
Tool Calling
60
Vision
60
Spatial Reasoning
50
Healthcare
50
Long Context
20
Structured Output
10

定价

输入价格$0.5 / 1M tokens
输出价格$3 / 1M tokens
混合价格(3:1)$1.125 / 1M tokens

速度

Tokens/秒197.0 tokens/s
首Token延迟0.81s
首回答延迟0.81s

可用提供商

(LS 内部计价单位)
提供商输入价格输出价格
Google500K3.0M

外部链接