44 models234 runs
🦀
Success rate by model
💰 Budget filter:Max $per run
Percentage of tasks completed successfully across standardized OpenClaw agent tests
Scores are graded via automated checks and LLM judge. How we benchmark·View all tasks
🦞
PinchBench
OpenClaw LLM Model Benchmarking
Success Rate by Model
Best score • 39 models
🦞
anthropic/claude-sonnet-4.686.9%
86.9%
🦀
anthropic/claude-opus-4.686.3%
86.3%
🦐
openai/gpt-5.486.0%
86.0%
nvidia/nemotron-3-super-120b-a12b85.6%
85.6%
anthropic/claude-opus-4.585.4%
85.4%
moonshotai/kimi-k2.584.8%
84.8%
qwen/qwen3.5-122b-a10b84.5%
84.5%
qwen/qwen3.5-plus-02-1584.1%
84.1%
z-ai/glm-584.1%
84.1%
anthropic/claude-sonnet-4.583.1%
83.1%
minimax/minimax-m2.182.2%
82.2%
deepseek/deepseek-v3.281.9%
81.9%
qwen/qwen3.5-397b-a17b81.7%
81.7%
stepfun/step-3.5-flash81.2%
81.2%
google/gemini-3.1-pro-preview81.1%
81.1%
anthropic/claude-sonnet-480.5%
80.5%
minimax/minimax-m2.580.5%
80.5%
z-ai/glm-4.5-air79.4%
79.4%
anthropic/claude-haiku-4.579.2%
79.2%
qwen/qwen3-coder-next79.1%
79.1%
mistralai/devstral-251278.6%
78.6%
openai/gpt-5-mini78.3%
78.3%
arcee-ai/trinity-large-preview:free77.7%
77.7%
qwen/qwen3.5-35b-a3b76.0%
76.0%
x-ai/grok-4.1-fast75.5%
75.5%
qwen/qwen3-max-thinking75.0%
75.0%
google/gemini-3-flash-preview74.9%
74.9%
openai/gpt-4o-mini70.9%
70.9%
qwen/qwen3.5-27b70.9%
70.9%
google/gemini-3-pro-preview70.7%
70.7%
deepseek/deepseek-chat69.5%
69.5%
openai/gpt-5-nano68.8%
68.8%
openai/gpt-4o64.7%
64.7%
mistralai/mistral-large-251263.8%
63.8%
google/gemini-2.5-flash61.9%
61.9%
google/gemini-2.5-pro61.4%
61.4%
openai/gpt-oss-20b57.4%
57.4%
openai/gpt-oss-120b54.0%
54.0%
qwen/qwen-2.5-7b-instruct40.3%
40.3%
pinchbench.com
Powered by Kilo Code
🦞
PinchBench
OpenClaw LLM Model Benchmarking
Success Rate Rankings
Best score • 39 models
All tasks and grading criteria are open source. Hover column headers for details.
pinchbench.com
Powered by Kilo Code
This leaderboard is for entertainment purposes only and should not be relied upon for making critical decisions.