Automated page. Written and translated automatically by AI and checked against the sources by software; no human editor. SI-IQ is an index built from public benchmark results, not an IQ test. How it works
Furniture Assembly: AI model results
Furniture Assembly is a Reasoning benchmark run by Epoch AI. 34 models in our data have a published result.
It is shown for reference and is not part of the SI-IQ index: fewer than 3 reference-group models.
- Evaluator
- Epoch AI
- Benchmark home page
- epoch.ai/benchmarks/furniture-assembly
Results
| 1 | Claude Opus 5.5 Anthropic | 83.3% | 144141–148 | 9 |
| 2 | GPT-6 Astra OpenAI | 80.0% | 150146–153 | 9 |
| 3 | GPT-6.1 Sol OpenAI | 80.0% | 148144–152 | 9 |
| 4 | Claude Sonnet 5.5 Anthropic | 75.0% | 140137–143 | 9 |
| 5 | Gemini 4 Argon Google DeepMind | 72.5% | 140137–143 | 9 |
| 6 | Claude Fable 5.1 Anthropic | 70.0% | 140136–143 | 8 |
| 7 | Claude Opus 5 Anthropic | 60.8% | 136133–140 | 9 |
| 8 | GPT-6 Sol OpenAI | 58.3% | 140137–144 | 9 |
| 9 | GPT-5.6 Sol OpenAI | 56.7% | 140137–143 | 9 |
| 10 | GPT-5.6 Terra OpenAI | 54.2% | 134131–137 | 9 |
| 11 | Muse Wonder Meta AI | 54.2% | 136133–139 | 4 too little data |
| 12 | Claude Haiku 5.5 Anthropic | 47.5% | 129126–132 | 9 |
| 13 | GPT-5.5 OpenAI | 44.2% | 138135–141 | 10 |
| 14 | GPT-6 Luna OpenAI | 44.2% | 130127–133 | 9 |
| 15 | Claude Opus 4.8 Anthropic | 42.5% | 132129–135 | 9 |
| 16 | GPT-5.6 Luna OpenAI | 42.5% | 130127–133 | 9 |
| 17 | Grok 4.6 xAI | 40.0% | 126123–129 | 9 |
| 18 | GPT-5.2 OpenAI | 38.3% | 125122–128 | 11 |
| 19 | GPT-5.4 OpenAI | 37.5% | 131128–133 | 10 |
| 20 | Claude Fable 5 Anthropic | 35.8% | 138134–141 | 9 |
| 21 | DeepSeek V4.1 Flashopen DeepSeek | 34.2% | 128125–131 | 9 |
| 22 | Kimi K3open Moonshot | 34.2% | 127124–130 | 9 |
| 23 | Claude Opus 4.7 Anthropic | 33.3% | 127124–130 | 10 |
| 24 | Gemini 3.8 Flash Google DeepMind | 31.7% | 130127–133 | 9 |
| 25 | Claude Opus 4.5 Anthropic | 28.3% | 115111–118 | 11 |
| 26 | Claude Opus 4.6 Anthropic | 28.3% | 123120–126 | 10 |
| 27 | Gemini 3.1 Pro Google DeepMind | 26.7% | 127124–130 | 10 |
| 28 | Gemini 3.7 Flash Google DeepMind | 26.7% | 131128–134 | 9 |
| 29 | Gemini 3.6 Flash Google DeepMind | 23.3% | 125122–128 | 9 |
| 30 | Grok 4.5 xAI | 22.5% | 124121–127 | 6 too little data |
| 31 | Kimi K2.6open Moonshot | 21.7% | 121118–124 | 10 |
| 32 | Grok 4.7 xAI | 20.8% | 123120–127 | 7 too little data |
| 33 | Mistral Large 4 Mistral AI | 20.8% | 122119–125 | 7 too little data |
| 34 | Qwen3.8 Max (0902) Alibaba | 20.0% | 127123–130 | 6 too little data |
Data sources: Aider polyglot leaderboard data (Apache-2.0); Arena leaderboard dataset on Hugging Face (CC BY 4.0); Berkeley Function Calling Leaderboard, Gorilla project (Apache-2.0); Epoch AI, ‘Capabilities & benchmarking’ (CC BY 4.0); LiteLLM, model_prices_and_context_window.json (MIT); Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter (licences and changes).