Totally An AdThe open source AI coding agent with 500+ models.Try Kilo
Quick Picks
anthropic/claude-opus-4.8-fastHighest average across benchmark runs.
nvidia/nemotron-3.5-bolt-74b-a9b-bf16Highest open-weights average across benchmark runs.
nvidia/nemotron-3.5-bolt-74b-a9b-bf16Highest average among open-weights models up to 125B total parameters.
inception/mercury-2Lowest observed complete benchmark runtime.
meta-llama/llama-4-scoutLowest observed non-zero benchmark run cost.
google/gemma-4-26b-a4b-itBest success percentage per dollar.
xiaomi/mimo-v2.5-proHighest score on code and DevOps tasks.
anthropic/claude-opus-4.8-fastHighest score on data analysis tasks.
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
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.
