# LiveClaw

> Personal-assistant agent tasks run end to end inside the OpenClaw agent harness.

Canonical page: https://benchlm.ai/benchmarks/liveclaw

- Category: [Agentic](/agentic)
- Last updated: October 29, 2026

## About LiveClaw

- Year: 2026
- Tasks: Assistant agent tasks
- Format: Task score
- Difficulty: Multi-step assistant work in a real harness
- Paper: [Introducing LFM3 (Liquid AI)](https://www.liquid.ai/blog/lfm3)

Liquid AI reports it for LFM3-1.7B (October 27, 2026), run with the OpenClaw agent at up to 8,192 tokens per turn. The launch post does not link a benchmark source, task count or version.

LiveClaw is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (1 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [LFM3-1.7B](/models/lfm3-1-7b) | LiquidAI | 45.60% |

## FAQ

### What does LiveClaw measure?

Personal-assistant agent tasks run end to end inside the OpenClaw agent harness.

### Which model scores highest on LiveClaw?

LFM3-1.7B by LiquidAI currently leads with a score of 45.60% on LiveClaw.

### How many models are evaluated on LiveClaw?

1 AI models have published results on LiveClaw in the BenchLM catalog.

### Does LiveClaw affect BenchLM's overall score?

Not directly. LiveClaw is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.