跳轉到主要內容

Qwen2.5-Omni-7B

Alibaba Cloud / Qwen TeamQwenOpen WeightApache 2.0 · Commercial OK

描述

Qwen2.5-Omni is the flagship end-to-end multimodal model in the Qwen series. It processes diverse inputs including text, images, audio, and video, delivering real-time streaming responses through text generation and natural speech synthesis using a novel Thinker-Talker architecture.

發布日期
2025-03-27
參數規模
7.0B
上下文長度
支援模態

能力雷達圖

50
general
80
coding
60
reasoning
26
science估算
0
agents
90
multimodal

Science 在缺少專門科學評測時使用推理能力代理估算。

排行榜排名

領域#排名分數來源
多模态榜52
74.0
LS

基準測試分數 (LLM Stats)

Audio

VocalSound93.9%自報
GiantSteps Tempo88.0%自報
MMAU Music69.2%自報
MMAU Sound67.9%自報
MMAU65.6%自報
MMAU Speech59.8%自報
OmniBench Music52.8%自報
CoVoST2 en-zh0.41 / 100自報
MusicCaps32.8%自報
Common Voice 150.08 / 100自報

Biology

GPQA30.8%自報

Code

HumanEval78.7%自報

Communication

VoiceBench Avg74.1%自報
MM-MT-Bench0.06 / 100自報

Creativity

Meld57.0%自報

Finance

MMLU-Pro47.0%自報

General

MBPP0.73 / 100自報
MMLU-Redux71.0%自報
MultiPL-E65.8%自報
MMStar64.0%自報
MME-RealWorld61.6%自報
MMMU59.2%自報
MMMU-Pro36.6%自報
LiveBench29.6%自報
NMOS0.05 / 100自報

Grounding

PointGrounding66.5%自報

Healthcare

CRPErelation76.5%自報

Image To Text

DocVQA95.2%自報
TextVQA84.4%自報
OCRBench_V257.8%自報

Language

FLEURS0.04 / 100自報

Long Context

EgoSchema68.6%自報

Math

GSM8k88.7%自報
MATH71.5%自報
MathVista67.9%自報
MathVision25.0%自報

Multimodal

ChartQA85.3%自報
AI2D83.2%自報
MMBench-V1.181.8%自報
VideoMME w sub.72.4%自報
MVBench70.3%自報
MuirBench59.2%自報
OmniBench56.1%自報

Spatial Reasoning

RealWorldQA70.3%自報

Vision

ODinW42.4%自報

AA 評測指數

暫無 AA 評測資料

LLM Stats 分類評分

Image To Text
90
Code
80
Spatial Reasoning
70
Video
70
Vision
70
Long Context
70
Math
60
Multimodal
60
Reasoning
60
Finance
50
General
50
Healthcare
50
Language
50
Legal
50
Biology
30
Chemistry
30
Physics
30
Communication
10
Speech To Text
0

定價

暫無定價資料

速度

暫無速度資料

可用提供商

(LS 內部計價單位)

暫無提供商資料

外部連結