PinchBench - Success Rate Leaderboard

Model	Success %	Score
🦞`moonshotai/kimi-k2.5`	83.5%	83.5%
🦀`anthropic/claude-opus-4.6`	81.7%	81.7%
🦐`qwen/qwen3.5-397b-a17b`	80.7%	80.7%
`z-ai/glm-5`	80.2%	80.2%
`x-ai/grok-4.1-fast`	80.0%	80.0%
`minimax/minimax-m2.5`	79.7%	79.7%
`anthropic/claude-sonnet-4.5`	78.4%	78.4%
`qwen/qwen3.5-35b-a3b`	78.4%	78.4%
`openai/gpt-5.4`	77.4%	77.4%
`qwen/qwen3.5-plus-02-15`	77.1%	77.1%
`minimax/minimax-m2.1`	77.0%	77.0%
`openai/gpt-5-mini`	76.3%	76.3%
`anthropic/claude-sonnet-4.6`	75.3%	75.3%
`nvidia/nemotron-3-super-120b-a12b:free`	75.0%	75.0%
`anthropic/claude-haiku-4.5`	74.8%	74.8%
`qwen/qwen3.5-27b`	74.7%	74.7%
`stepfun/step-3.5-flash`	74.1%	74.1%
`qwen/qwen3.5-122b-a10b`	74.0%	74.0%
`google/gemini-3.1-pro-preview`	73.3%	73.3%
`qwen/qwen3-max-thinking`	71.5%	71.5%
`openai/gpt-4o-mini`	71.5%	71.5%
`z-ai/glm-4.5-air`	70.8%	70.8%
`mistralai/devstral-2512`	70.7%	70.7%
`deepseek/deepseek-v3.2`	69.7%	69.7%
`anthropic/claude-opus-4.5`	69.2%	69.2%
`google/gemini-3-flash-preview`	67.4%	67.4%
`openai/gpt-4o`	66.5%	66.5%
`google/gemini-2.5-pro`	65.5%	65.5%
`google/gemini-2.5-flash`	65.3%	65.3%
`arcee-ai/trinity-large-preview:free`	64.2%	64.2%
`openai/gpt-5-nano`	62.0%	62.0%
`mistralai/mistral-large-2512`	56.2%	56.2%
`openai/gpt-oss-20b`	55.7%	55.7%
`deepseek/deepseek-chat`	55.1%	55.1%

All tasks and grading criteria are open source. Hover column headers for details.