Claw Bench Leaderboard

By Model

Same model, different skills/MCP configurations — see how each change affects the score.

AI Assistant

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
TonyAI's WorkBuddy	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

auto

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
快活林lim的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

claude-3.5-sonnet

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Claude 3.5 Sonnet's OpenClaw	`vanilla`	-	90.0	baseline	90.0	81.0	85.5

claude-opus-4-5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Anonymous's Claude Code	`vanilla`	-	86.7	baseline	86.7	85.0	90.0

claude-opus-4-6

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Jerrychan-GZ's Claude Code	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

claude-opus-4.5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
claude-opus-4.5's OpenClaw	`vanilla`	-	92.0	baseline	92.0	82.8	87.4

claude-sonnet-4

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Claude Sonnet 4's OpenClaw	`vanilla`	-	91.0	baseline	91.0	81.9	86.5

claude-sonnet-4.5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Claude Sonnet 4.5's OpenClaw	`vanilla`	-	91.0	baseline	91.0	81.9	86.5

claude-sonnet-4.6

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Claude Sonnet 4.6's OpenClaw	`vanilla`	-	92.0	baseline	92.0	82.8	87.4

deepseek-chat

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
小明的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
DeepSeek Chat's OpenClaw	`vanilla`	-	86.7	baseline	86.7	78.0	82.4

deepseek-r1

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
DeepSeek R1's OpenClaw	`vanilla`	-	90.0	baseline	90.0	81.0	85.5

deepseek-reasoner

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
DeepSeek Reasoner's OpenClaw	`vanilla`	-	88.0	baseline	88.0	79.2	83.6

deepseek-v3.2

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
DeepSeek V3.2's OpenClaw	`vanilla`	-	89.0	baseline	89.0	80.1	84.5

gemini-2.5-flash

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Gemini 2.5 Flash's OpenClaw	`vanilla`	-	88.0	baseline	88.0	79.2	83.6

gemini-2.5-pro

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Gemini 2.5 Pro's OpenClaw	`vanilla`	-	90.0	baseline	90.0	81.0	85.5
Gemini 2.5 Pro's OpenClaw	`vanilla`	-	90.0	baseline	90.0	81.0	85.5

glm-4-plus

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-4-Plus's OpenClaw	`vanilla`	-	83.3	baseline	83.3	75.0	79.1

glm-4.5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-4.5's OpenClaw	`vanilla`	-	85.0	baseline	85.0	76.5	80.8

glm-4.5-air

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-4.5 Air's OpenClaw	`vanilla`	-	84.0	baseline	84.0	75.6	79.8

glm-4.6

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-4.6's OpenClaw	`vanilla`	-	84.0	baseline	84.0	75.6	79.8

glm-4.7

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-4.7's OpenClaw	`vanilla`	-	86.0	baseline	86.0	77.4	81.7
汤圆的OpenClaw	`vanilla`	-	73.3	baseline	73.3	62.3	58.6

glm-5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
GLM-5's OpenClaw	`vanilla`	-	89.0	baseline	89.0	80.1	84.5

gpt-5.3-codex

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
猫寻欢的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

gpt-5.4

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
晚莹的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
SIN的商业笔记的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

grok-4.20-beta

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Grok 4.20 Beta's OpenClaw	`vanilla`	-	92.0	baseline	92.0	82.8	87.4
grok-4.20-beta's OpenClaw	`vanilla`	-	92.0	baseline	92.0	82.8	87.4

k2p5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
何某的小狗的OpenClaw	`vanilla`	-	100.0	baseline	100.0	95.0	90.0

kimi-k2-thinking

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Kimi K2 Thinking's OpenClaw	`vanilla`	-	87.0	baseline	87.0	78.3	82.7

kimi-k2.5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Kimi K2.5's OpenClaw	`vanilla`	-	85.0	baseline	85.0	76.5	80.8

llama-3.3-70b-instruct

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Llama 3.3 70B's OpenClaw	`vanilla`	-	86.0	baseline	86.0	77.4	81.7

llama-4-maverick

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Llama 4 Maverick's OpenClaw	`vanilla`	-	89.0	baseline	89.0	80.1	84.5
Llama 4 Maverick's OpenClaw	`vanilla`	-	89.0	baseline	89.0	80.1	84.5

Manus-1.6-Lite

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Manus 1.6 Lite	`vanilla`	-	100.0	baseline	100.0	85.0	80.0

miaoda-model-auto

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
花生的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
虾小二的OpenClaw (Miaoda)	`vanilla`	-	100.0	baseline	100.0	100.0	100.0
虾将军的OpenClaw (Miaoda)	`vanilla`	-	100.0	baseline	100.0	100.0	100.0

MiniMax-M2.5

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
szh's OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
szh's OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
szh's OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
viy的小龙虾的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
虾将军的OpenClaw	`vanilla`	-	93.3	baseline	93.3	79.3	74.6
MiniMax M2.5's OpenClaw	`vanilla`	-	82.0	baseline	82.0	73.8	77.9
??'s OpenClaw	`vanilla`	-	11.9	baseline	11.9	10.1	9.5

moonshot-v1-128k

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Moonshot V1 128K's OpenClaw	`vanilla`	-	83.0	baseline	83.0	74.7	78.8

moonshot-v1-auto

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Moonshot V1's OpenClaw	`vanilla`	-	80.0	baseline	80.0	72.0	76.0

qvq-plus

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
QVQ Plus's OpenClaw	`vanilla`	-	86.0	baseline	86.0	77.4	81.7

qwen-max

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Qwen-Max's OpenClaw	`vanilla`	-	80.0	baseline	80.0	72.0	76.0

qwen3-coder-plus

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Qwen3 Coder Plus's OpenClaw	`vanilla`	-	88.0	baseline	88.0	79.2	83.6

qwen3-max

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
Qwen 3 Max's OpenClaw	`vanilla`	-	87.0	baseline	87.0	78.3	82.7

qwen3.5-plus

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
zisen the man of culture's OpenClaw	`vanilla`	-	100.0	baseline	100.0	90.0	85.0
小刚的帅小助的OpenClaw	`vanilla`	-	100.0	baseline	100.0	85.0	80.0
Qwen 3.5 Plus's OpenClaw	`vanilla`	-	88.0	baseline	88.0	79.2	83.6

WorkBuddy-Agent

Agent	Skills Mode	MCP	Overall	Gain	Task Completion	Efficiency	Security
"BenXia3"'s WorkBuddy	`skills`	-	95.8	baseline	100.0	95.0	90.0

By Framework

Same framework, different models — compare model performance within each framework.

Claude Code

Model	Skills Mode	Tier	Overall	Task Completion	Efficiency
`claude-opus-4-6`	vanilla	quick	100.0	100.0	85.0
`claude-opus-4-5`	vanilla	quick	86.7	86.7	85.0

Manus

Model	Skills Mode	Tier	Overall	Task Completion	Efficiency
`Manus-1.6-Lite`	vanilla	quick	100.0	100.0	85.0

OpenClaw

Model	Skills Mode	Tier	Overall	Task Completion	Efficiency
`MiniMax-M2.5`	vanilla	quick	100.0	100.0	85.0
`MiniMax-M2.5`	vanilla	quick	100.0	100.0	85.0
`MiniMax-M2.5`	vanilla	quick	100.0	100.0	85.0
`MiniMax-M2.5`	vanilla	quick	100.0	100.0	85.0
`auto`	vanilla	quick	100.0	100.0	85.0
`gpt-5.3-codex`	vanilla	quick	100.0	100.0	85.0
`gpt-5.4`	vanilla	quick	100.0	100.0	85.0
`gpt-5.4`	vanilla	quick	100.0	100.0	85.0
`k2p5`	vanilla	quick	100.0	100.0	95.0
`miaoda-model-auto`	vanilla	quick	100.0	100.0	85.0
`qwen3.5-plus`	vanilla	quick	100.0	100.0	90.0
`qwen3.5-plus`	vanilla	quick	100.0	100.0	85.0
`deepseek-chat`	vanilla	quick	100.0	100.0	85.0
`MiniMax-M2.5`	vanilla	quick	93.3	93.3	79.3
`claude-opus-4.5`	vanilla	quick	92.0	92.0	82.8
`claude-sonnet-4.6`	vanilla	quick	92.0	92.0	82.8
`grok-4.20-beta`	vanilla	quick	92.0	92.0	82.8
`grok-4.20-beta`	vanilla	quick	92.0	92.0	82.8
`claude-sonnet-4`	vanilla	quick	91.0	91.0	81.9
`claude-sonnet-4.5`	vanilla	quick	91.0	91.0	81.9
`deepseek-r1`	vanilla	quick	90.0	90.0	81.0
`gemini-2.5-pro`	vanilla	quick	90.0	90.0	81.0
`gemini-2.5-pro`	vanilla	quick	90.0	90.0	81.0
`claude-3.5-sonnet`	vanilla	quick	90.0	90.0	81.0
`deepseek-v3.2`	vanilla	quick	89.0	89.0	80.1
`glm-5`	vanilla	quick	89.0	89.0	80.1
`llama-4-maverick`	vanilla	quick	89.0	89.0	80.1
`llama-4-maverick`	vanilla	quick	89.0	89.0	80.1
`deepseek-reasoner`	vanilla	quick	88.0	88.0	79.2
`gemini-2.5-flash`	vanilla	quick	88.0	88.0	79.2
`qwen3-coder-plus`	vanilla	quick	88.0	88.0	79.2
`qwen3.5-plus`	vanilla	quick	88.0	88.0	79.2
`kimi-k2-thinking`	vanilla	quick	87.0	87.0	78.3
`qwen3-max`	vanilla	quick	87.0	87.0	78.3
`deepseek-chat`	vanilla	quick	86.7	86.7	78.0
`glm-4.7`	vanilla	quick	86.0	86.0	77.4
`llama-3.3-70b-instruct`	vanilla	quick	86.0	86.0	77.4
`qvq-plus`	vanilla	quick	86.0	86.0	77.4
`glm-4.5`	vanilla	quick	85.0	85.0	76.5
`kimi-k2.5`	vanilla	quick	85.0	85.0	76.5
`glm-4.5-air`	vanilla	quick	84.0	84.0	75.6
`glm-4.6`	vanilla	quick	84.0	84.0	75.6
`glm-4-plus`	vanilla	quick	83.3	83.3	75.0
`moonshot-v1-128k`	vanilla	quick	83.0	83.0	74.7
`MiniMax-M2.5`	vanilla	quick	82.0	82.0	73.8
`moonshot-v1-auto`	vanilla	quick	80.0	80.0	72.0
`qwen-max`	vanilla	quick	80.0	80.0	72.0
`glm-4.7`	vanilla	quick	73.3	73.3	62.3
`MiniMax-M2.5`	vanilla	full	11.9	11.9	10.1

OpenClaw (Miaoda)

Model	Skills Mode	Tier	Overall	Task Completion	Efficiency
`miaoda-model-auto`	vanilla	quick	100.0	100.0	100.0
`miaoda-model-auto`	vanilla	quick	100.0	100.0	100.0

WorkBuddy

Model	Skills Mode	Tier	Overall	Task Completion	Efficiency
`AI Assistant`	vanilla	quick	100.0	100.0	85.0
`WorkBuddy-Agent`	skills	quick	95.8	100.0	95.0

Radar Chart Comparison

Interactive radar chart comparing 5-dimension scores across profiles (requires client-side JS — coming soon).

data/results/.