Qwen3.5 397B A17B (Reasoning)
AlibabaQwen
发布日期
2026-02-16
参数规模
—
上下文长度
262K
支持模态
audio, image, text, video
能力雷达图
32
general
47
coding
89
reasoning
60
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
基准测试分数 (LLM Stats)
3d
SUNRGBD
0.36 / 100自报
Hypersim
0.13 / 100自报
Agents
GDPval-AA
985.00 / 3000自报
t2-bench
79.5%自报
BFCL-V4
72.2%自报
AndroidWorld_SR
66.4%自报
BrowseComp
63.8%自报
FullStackBench en
62.6%自报
WideSearch
60.5%自报
FullStackBench zh
58.7%自报
OSWorld-Verified
58.0%自报
TIR-Bench
53.2%自报
Terminal-Bench 2.0
49.4%自报
VITA-Bench
33.6%自报
DeepPlanning
24.1%自报
Biology
GPQA
86.6%自报
Chemistry
SuperGPQA
67.1%自报
Code
SWE-Bench Verified
72.0%自报
Communication
Multi-Challenge
61.5%自报
Embodied
EmbSpatialBench
0.84 / 100自报
Finance
MMLU-Pro
86.7%自报
MMLU-ProX
82.2%自报
General
MMLU-Redux
94.0%自报
IFEval
93.4%自报
C-Eval
91.9%自报
Global PIQA
88.4%自报
MAXIFE
87.9%自报
MMMLU
86.7%自报
MMMU
83.9%自报
MMStar
82.9%自报
Include
82.8%自报
LiveCodeBench v6
78.9%自报
MMMU-Pro
76.9%自报
IFBench
76.1%自报
SimpleVQA
0.62 / 100自报
LongBench v2
60.2%自报
NOVA-63
58.6%自报
Grounding
RefCOCO-avg
0.91 / 100自报
ScreenSpot Pro
70.4%自报
RefSpatialBench
0.69 / 100自报
Healthcare
VideoMMMU
82.0%自报
SlakeVQA
81.6%自报
MedXpertQA
67.3%自报
PMC-VQA
63.3%自报
Image To Text
OCRBench
92.1%自报
Language
LingoQA
80.8%自报
WMT24++
78.3%自报
Long Context
MLVU
87.3%自报
LVBench
74.4%自报
AA-LCR
66.9%自报
MMLongBench-Doc
0.59 / 100自报
Math
HMMT 2025
91.4%自报
HMMT25
90.3%自报
MathVista-Mini
87.4%自报
MathVision
86.2%自报
DynaMath
85.9%自报
CodeForces
0.85 / 3000自报
PolyMATH
68.9%自报
Humanity's Last Exam
47.5%自报
Multimodal
VLMsAreBlind
96.7%自报
AI2D
93.3%自报
V*
93.2%自报
MMBench-V1.1
92.8%自报
OmniDocBench 1.5
89.8%自报
VideoMME w sub.
87.3%自报
VideoMME w/o sub.
83.9%自报
CC-OCR
81.8%自报
CharXiv-R
77.2%自报
MVBench
76.6%自报
MMVU
74.7%自报
BabyVision
40.2%自报
ZEROBench-Sub
0.36 / 100自报
Nuscene
15.4%自报
ZEROBench
0.09 / 100自报
Reasoning
CountBench
0.97 / 100自报
BrowseComp-zh
69.9%自报
Hallusion Bench
67.6%自报
ERQA
62.0%自报
Seal-0
44.1%自报
OJBench
39.5%自报
Spatial Reasoning
RealWorldQA
85.1%自报
Vision
ODinW
44.5%自报
AA 评测指数
Coding Index48.2
Intelligence Index33.7
Tau21.0
Gpqa0.9
Ifbench0.8
Lcr0.7
Terminalbench V2 10.5
Scicode0.4
Terminalbench Hard0.4
Hle0.3
Tau Banking0.1
LLM Stats 分类评分
Legal100
Finance100
Agents76
General46
Reasoning19
Biology90
Image To Text80
Instruction Following80
Language80
Math80
Physics80
Structured Output80
Embodied80
Grounding80
Healthcare80
Chemistry80
Text-to-image80
Video80
Long Context70
Multimodal70
Spatial Reasoning70
Frontend Development70
Economics70
Vision70
Search60
Code60
Communication60
Tool Calling60
Spatial20
3d20
定价
输入价格$0.6 / 1M tokens
输出价格$3.6 / 1M tokens
混合价格(3:1)$1.35 / 1M tokens
速度
Tokens/秒51.7
首Token延迟1.76s
首回答延迟63.34s
供应商价格排行
供应商价格排行
18 个供应商
最便宜: OrcaRouter最贵: Venice AI
供应商输入输出
1OrcaRouter最便宜
$0.172
$1.032
2SiliconFlow (China)
$0.29
$1.74
3OpenRouter
$0.385
$2.45
4Chutes
$0.39
$2.34
5Kilo Gateway
$0.39
$2.34
6Wafer
$0.43
$2.6
7Deep Infra
$0.45
$3
8Alibaba主要
$0.6
$3.6
9NanoGPT
$0.6
$3.6
10Hugging Face
$0.6
$3.6
11NovitaAI
$0.6
$3.6
12Mixlayer
$0.6
$3.6
13Nebius Token Factory
$0.6
$3.6
14Synthetic
$0.6
$3
15LLM Gateway
$0.6
$3.6
16Together AI
$0.6
$3.6
17Neuralwatt
$0.69
$4.14
18Venice AI
$0.75
$4.5
比较该模型在不同 API 供应商之间的定价。