DeepSeek V3.2 Speciale
DeepSeekDeepSeek开源权重MIT · 商用许可
描述
DeepSeek-V3.2-Speciale is a specialized variant of DeepSeek-V3.2, optimized for enhanced performance on specific tasks.
发布日期
2025-12-01
参数规模
685.0B
上下文长度
—
支持模态
text
能力雷达图
43
general
79
coding
95
reasoning
60
science
50
agents
0
multimodal
排行榜排名
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Agents
t2-bench
80.3%自报
Terminal-Bench 2.0Stanford × Laude Institute (2026)
46.4%自报
Toolathlon
35.2%自报
Code
SWE-Bench Verified
73.1%自报
Math
HMMT 2025
99.2%自报
AIME 2025
96.0%自报
CodeForces
0.90 / 3000自报
Humanity's Last Exam
30.6%自报
AA 评测指数
(Artificial Analysis)Math Index(Artificial Analysis)96.7
Intelligence Index(Artificial Analysis)22.6
Aime 25(MAA (Mathematical Association of America))1.0
Livecodebench(UC Berkeley + MIT + Cornell (2024))0.9
Gpqa(NYU + Cohere + Anthropic (2023))0.9
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))0.9
Lcr(Artificial Analysis)0.6
Ifbench(Google Research (2023))0.6
Scicode(UIUC + Argonne National Lab (2024))0.4
Terminalbench Hard(Stanford × Laude Institute (2026))0.3
Hle(Center for AI Safety + Scale AI (2025))0.3
Tau2(Sierra + U Toronto + Vector Institute (2025))0.0
LLM Stats 分类评分
(LLM Stats (zeroeval))Math80
Frontend Development70
General70
Reasoning60
Code60
Agents50
Tool Calling50
Vision30
定价
输入价格免费
输出价格免费
混合价格(3:1)免费
速度
Tokens/秒0.0
首Token延迟0.00s
首回答延迟0.00s
供应商价格排行
暂无提供商数据