Command A+
CohereCommand开源权重Apache 2.0 · 商用许可
描述
Command A+ is Cohere's open-source sparse Mixture-of-Experts model for enterprise agentic, multilingual, multimodal, and reasoning-heavy tasks. It has 218B total parameters with 25B active parameters, supports text and image inputs, conversational tool use, generated thinking, 48 languages, a 128K input context, and up to 64K output tokens.
发布日期
2026-05-20
参数规模
218.0B
上下文长度
256K
支持模态
text
能力雷达图
20
general
29
coding
76
reasoning
50
science
80
agents
0
multimodal
排行榜排名
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Agents
Cohere Agentic Question Answering
65.0%自报
Cohere Memory Usage Quality
54.0%自报
Communication
Tau2 Telecom
85.0%自报
Data Analysis
Cohere Data Analysis
45.0%自报
Instruction Following
IFBench
74.0%自报
Language
WMT24++
81.0%自报
Math
AIME 2025
90.0%自报
MT-AIME 2025
86.0%自报
MathVista
80.6%自报
Multimodal
MMMU
75.1%自报
Reasoning
CharXiv-D
88.0%自报
CharXiv-R
52.7%自报
SciCode
38.0%自报
Terminal-Bench Hard
25.0%自报
Vision
MMMU-Pro
63.0%自报
AA 评测指数
(Artificial Analysis)Tau2(Sierra + U Toronto + Vector Institute (2025))80.7
Gpqa(NYU + Cohere + Anthropic (2023))76.1
Ifbench(Google Research (2023))73.9
Lcr(Artificial Analysis)48.7
Scicode(UIUC + Argonne National Lab (2024))37.8
Coding Index(Artificial Analysis)27.8
Terminalbench Hard(Stanford × Laude Institute (2026))25.0
Terminalbench V2 122.8
Intelligence Index(Artificial Analysis)22.8
Hle(Center for AI Safety + Scale AI (2025))12.0
Tau Banking6.0
LLM Stats 分类评分
(LLM Stats (zeroeval))Structured Output90
Language80
Healthcare80
Communication80
Tool Calling80
Math70
Multimodal70
Reasoning70
Instruction Following70
General70
Vision70
Physics40
Biology40
Chemistry40
Code40
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239.9
首Token延迟0.18s
首回答延迟8.51s
供应商价格排行
供应商价格排行
2 个供应商
最便宜: NanoGPT最贵: Cohere
供应商输入输出
1NanoGPT最便宜
$2.5
$10
2Cohere
$2.5
$10
比较该模型在不同 API 供应商之间的定价。