83%
best furniture-assembly score, Claude Opus 5.5 (max) (Anthropic), of 36 models
What the chart shows
Epoch AI’s furniture-assembly benchmark asks a model to work out how to assemble real furniture, a test of the physical reasoning robots need. Claude Opus 5.5 (max) (Anthropic) scored 83%, ahead of GPT-6.1 Sol (max) (80%).
It tests the model’s planning, not a robot’s hands: no physical robot is involved.
Key numbers
| Measure | Value |
|---|---|
| Claude Opus 5.5 (max) (Anthropic) | 83% |
| GPT-6.1 Sol (max) (OpenAI) | 80% |
| GPT-6 Astra (max) (OpenAI) | 80% |
| GPT-6.1 Terra (max) (OpenAI) | 79% |
| Claude Sonnet 5.5 (max) (Anthropic) | 75% |
| Gemini 4 Argon (high) (Google DeepMind) | 72% |
How to read it
- Epoch AI’s own runs; each model’s best score across scorers is shown.
- All models shown were tested in 2026.
Source: Epoch AI, 2026: Benchmarking Hub, read 3 November 2026. Chart and text by AI Adoption.
Numbers checked against the published data on 3 November 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 3 November 2026.
Download the data (CSV) · Cite this chart
Discussion