Skip to content

PinchBench

An OpenClaw agent benchmark from Kilo that measures successful task completion across standardized real-world agent workflows.

7 models

Higher is better.

  1. 1Pokee-Isaac 28BPokee AI95.7
  2. 2Nemotron 3 UltraNVIDIA90.0
  3. 3Nemotron 3.5 Bolt 74B-A9BNVIDIA87.8
  4. 4Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA83.4
  5. 5LFM2.5-24B-A2BLiquid AI83.2
  6. 6LLaDA2.2-flashInclusionAI81.7
  7. 7LFM2.5-2.6BLiquid AI68.2
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Average success rate from official runs
Tasks
23 OpenClaw agent tasks
Authors
Kilo Code (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.