Benchmarks / Agentic
Butter-Bench
LLMs orchestrate a household robot through six 'pass the butter' subtasks; average task completion rate.
Butter-Bench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | high | 40.0% |
| 2 | Opus 4.1 | high | 37.0% |
| 3 | GPT-5 | high | 30.0% |
| 4 | Grok 4 | high | 23.0% |
| 5 | Llama 4 Maverick | medium | 7.0% |