Seed 2.0 Mini
ByteDanceProprietary
描述
Built for low-latency, high-concurrency, cost-sensitive use cases, with flexible deployment, four-tier thinking, and multimodal
发布日期
2026-02-17
参数规模
—
上下文长度
262K
支持模态
image, text, video
能力雷达图
21
general
10
coding
100
reasoning
68
science估算
10
agents
70
multimodal
缺少专门科学评测时，Science 由 LLM Stats 科学得分或推理能力估算。
排行榜排名
| 领域 | #排名 | 分数 | 来源 |
|---|---|---|---|
| 多模态榜 | 124 | 34.0 | LS |
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Chat
Multi-Challenge
61.1%自报
Chemistry
SuperChem
16.2%自报
Code
CL-bench
14.8%自报
Coding
AetherCode
29.8%自报
Factuality
SimpleQA Verified
18.9%自报
General
LongFact Concepts
91.4%自报
LongFact Objects
87.0%自报
Healthcare
HealthBench
30.0%自报
HealthBench Hard
15.3%自报
Language
COLLIE
91.2%自报
MMLU-Pro
83.6%自报
MMMLU
81.6%自报
Long Context
MRCR v2 (8-needle)
51.4%自报
Math
CodeForces
1644.00 / 3000自报
AIME 2025
87.0%自报
AIME 2026
86.7%自报
MathVista-Mini
85.5%自报
MathVision
78.1%自报
Beyond AIME
69.0%自报
DynaMath
58.9%自报
MathArena Apex
4.2%自报
Multimodal
VideoMMMU
80.6%自报
MMMU
79.7%自报
Physics
PHYBench
56.0%自报
Reasoning
Graphwalks parents <128k
93.0%自报
CharXiv-D
91.9%自报
Global PIQA
89.2%自报
FRAMES
80.5%自报
GPQANYU + Cohere + Anthropic (2023)
79.0%自报
CharXiv-R
70.8%自报
LiveCodeBench v6
64.1%自报
Graphwalks BFS <128k
64.1%自报
SuperGPQA
61.6%自报
FActScoreMin et al. (NYU/UW, 2023)
50.4%自报
FrontierScience Olympiad
44.0%自报
ARC-AGI
29.8%自报
Humanity's Last Exam (no tools, text-only)
13.3%自报
FrontierScience Research
3.3%自报
ArcAGI2
2.3%自报
ARC-AGI v2
1.5%自报
Video
LiveSports-3K
73.3%自报
OVOBench
70.4%自报
VideoSimpleQA
67.7%自报
OVBench
60.1%自报
VideoHolmes
58.6%自报
CrossVid
58.6%自报
ContPhy
55.9%自报
Minerva
54.7%自报
TOMATO
47.4%自报
Vision
CountBench
0.95 / 100自报
VLMsAreBlind
93.1%自报
TempCompass
83.7%自报
RealWorldQA
81.6%自报
VideoMME w sub.
81.2%自报
MMStar
79.1%自报
MuirBench
78.0%自报
LongVideoBench
74.8%自报
BLINK
73.4%自报
MMMU-Pro
71.4%自报
TVBench
70.5%自报
MMVU
69.0%自报
DUDE
68.9%自报
SimpleVQA
0.69 / 100自报
LVBench
66.6%自报
ChartQAPro
65.2%自报
Hallusion Bench
65.1%自报
MotionBench
64.4%自报
OCRBench_V2
58.5%自报
VLMsAreBiased
58.4%自报
TreeBench
57.3%自报
EMMA
57.0%自报
ERQA
56.3%自报
HiPhO
55.8%自报
RefSpatialBench
0.56 / 100自报
MTVQA
50.6%自报
MMLongBench-Doc
0.49 / 100自报
WorldVQA
47.6%自报
VisuLogic
0.40 / 100自报
BabyVision
38.7%自报
ZEROBench-Sub
0.36 / 100自报
VisFactor
23.6%自报
OmniDocBench 1.5
11.0%自报
ZEROBench
0.07 / 100自报
AA 评测指数
(Artificial Analysis)暂无 AA 评测数据
LLM Stats 分类评分
(LLM Stats (zeroeval))Math100
Reasoning28
General21
Language90
Writing90
Search80
Biology80
Legal70
Physics70
Finance70
3d70
Chat60
Image To Text60
Long Context60
Multimodal60
Spatial Reasoning60
Grounding60
Healthcare60
Communication60
Economics60
Video60
Vision60
Structured Output50
Chemistry50
Knowledge40
Science30
Coding30
Factuality20
Agents10
Code10
定价
输入价格$0 / 1M tokens
输出价格$0 / 1M tokens
混合价格(3:1)$0 / 1M tokens
速度
暂无速度数据
供应商价格排行
供应商价格排行
5 个供应商
最便宜: ByteDance最贵: Requesty
供应商输入输出
1ByteDance主要
$0
$0
2DeepInfra
$0
$0
3OpenRouter
$0.1
$0.4
4Kilo Gateway
$0.1
$0.4
5Requesty
$0.1
$0.4
比较该模型在不同 API 供应商之间的定价。