Category:
The best models for your OpenClaw agent.
Quick Picks
Best AI models for common use cases
Best Overall
93.5%anthropic/claude-opus-4.8-fastAverage Score93.5% overall · $159.60
Highest average across benchmark runs.
Best Open-Weights
89.9%nvidia/nemotron-3-ultra-550b-a55bAverage Score89.9% overall · FREE
Highest open-weights average across benchmark runs.
Fastest
11.0sinception/mercury-2Best Time100.0% overall · FREE
Lowest observed complete benchmark runtime.
Best Budget
$0.243meta-llama/llama-4-scoutBest Cost3.2% overall · $0.243
Lowest observed non-zero benchmark run cost.
Best Value
173.1google/gemma-4-26b-a4b-itValue Score77.1% overall · $0.445
Best success percentage per dollar.
Best for Code
98.0%xiaomi/mimo-v2.5-proCoding Score89.5% overall · $11.39
Highest score on code and DevOps tasks.
Best for Data
97.1%anthropic/claude-opus-4.8-fastData Score94.5% overall · $159.60
Highest score on data analysis tasks.
🦀
Success rate by model
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
Totally An Ad
The open source AI coding agent with 500+ models.
Hosting and inference for PinchBench is sponsored by Kilo, so we totally hope you try kilo.ai so we can keep the lights on around here.
