LiveClaw
A benchmark for evaluating on-device agent capabilities, focusing on assessing model performance in local environments across tool calling, multi-step tasks, and structured output.
Timeline (last 90 days)
LFM2.5-24B-A2B scored 48.09 on the LiveClaw benchmark.
A LiveClaw score of 48.09 reflects the model's performance on a specific agent evaluation set, and does not directly equate to general reasoning or coding ability.
When judging on-device agent models, agent-specific benchmarks (such as LiveClaw) better reflect real usability than general question-answering leaderboards.
All Facts (3)
LFM2.5-24B-A2B scored 48.09 on the LiveClaw benchmark.
50%UnverifiedA LiveClaw score of 48.09 reflects the model's performance on a specific agent evaluation set, and does not directly equate to general reasoning or coding ability.
50%UnverifiedWhen judging on-device agent models, agent-specific benchmarks (such as LiveClaw) better reflect real usability than general question-answering leaderboards.
50%